United States v. Bain

874 F.3d 1, 2017 WL 4563821, 2017 U.S. App. LEXIS 20032

• Court: Court of Appeals for the First Circuit
• Decided: October 13, 2017
• Docket: 16-1140P
• Status: Published

Opinion

KAYATTA, Circuit Judge.

The police arrested Yrvens Bain after he emerged from a multi-family building in Malden, Massachusetts. During the search incident to that arrest, they found a set of keys in his possession. The police tried these keys on the front door of the multifamily building and on the doors to three apartments inside—one on the first floor, two on the second floor. The keys opened the door to one of the units on the second floor. The police included this information in an application for a warrant to search that unit. The warrant issued, and the search produced a firearm and over twenty-six grams of heroin mixed with fentanyl. Bain moved to suppress that evidence. He argued, among other things, that the officers conducted an unlawful search by turning his key in the locks to identify the unit to search, and that there was no probable cause to issue a warrant to search the unit without that identification.

The district court denied Bain’s motion. The court also subsequently denied his motion in limine to exclude a credit-card-making machine found during the search. At trial, a jury convicted Bain on two counts of distribution of heroin, see 21 U.S.C. § 841(a)(1), one count of possessing heroin with intent to distribute, see id., and one count of possessing a firearm and ammunition after a conviction for a felony punishable by over one year in prison, see 18 U.S.C. § 922(g)(1). At sentencing, the district court applied the fifteen-year mandatory minimum sentence under the Armed Career Criminal Act (ACCA), see id. § 924(e). Bain appeals the rulings on his motion to suppress, his motion in li-mine, and his sentence. While we agree with Bain that the officers conducted an unlawful search by testing the key in the lock of the unit in which he was staying, we nevertheless affirm the denial of the motion to suppress because in searching the apartment the officers relied in good faith on the intervening warrant. We also affirm Bain’s conviction and sentence.

I.

A.

Brian Connerney, a detective in the Arlington, Massachusetts Police Department and a Task Force Agent with the Drug Enforcement Administration (DEA), signed both affidavits supporting the search warrant application. 1 We describe the relevant information contained in those affidavits.

In early 2014, the DEA began investigating Bain. At the time, Bain had four prior convictions for “drug trafficking offenses.” All four involved cocaine and the most recent had occurred in 2007. By 2014, Bain’s six-year prison sentence for that conviction had ended and his five-year probation term had just begun.

As part of the investigation, a cooperating witness made two controlled buys from Bain. In connection with both controlled buys, officers searched the cooperating witness beforehand, provided him with a recording device and the cash used to purchase the drugs, and retrieved the drugs from him afterwards.

The cooperating witness made the first controlled buy on February 26, 2014. After texting Bain to request $100 worth of heroin, the cooperating witness picked Bain up at an apartment complex in Waltham, Massachusetts, next to which Bain’s car was parked. 2 The cooperating witness paid $100 in cash for a baggie containing 0.80 grams of a mixture of heroin and fentanyl.

Three days later, police responded to a report of a fight in progress at the same apartment complex in Waltham. Bain had been living in his brother’s apartment in that complex, the two men had been in an argument, and Bain had punched his brother. These events precipitated an assault and battery charge against Bain, which landed him in police custody until March 17. When Bain was released, he informed his probation officer that he planned to live at his mother’s residence in Arlington. But Connerney never saw Bain at that residence and never saw his car parked outside.

The day after Bain’s release, he texted the cooperating witness, saying, “I was away for awhile but now I’m back hit me up everything’s good.” After the cooperating witness again requested $100 worth of heroin, Bain directed him to “Waite st.” in “Malden.” Bain once again got into the cooperating witness’s car and sold him a baggie containing heroin for $100 in cash. The cooperating witness drove around the block and let Bain out of the car on Webster Street.

On several subsequent occasions, Con-nerney and other officers involved in the investigation observed Bain’s car parked on Webster Street in Malden, near the intersection with Laurel Street. On March 28 at 2:35 P.M., officers in the Malden Police Department observed Bain park on Webster Street, walk to Laurel Street, and enter 131 Laurel Street. Roughly ten hours later, at 12:30 A.M. on March 29, Connerney observed Bain’s car still parked on Webster Street.’

Two days later, with a signed criminal complaint in hand, officers went to 131 Laurel Street to arrest Bain. They saw Bain emerge through the front door of 131 Laurel Street, walk to his .car around the corner on Webster Street, and get inside. As agents approached the car, Bain locked his doors and swallowed something. Agents removed him from the car, placed him under arrest, conducted a search incident to arrest, and seized, credit cards and a set of keys from his person,

The agents used the keys they had seized from Bain to open the front door of 131 Laurel Street. Connerney described the building located at 131 Laurel Street as follows:

131 Laurel Street is a two and a half story home located near the intersection of Laurel Street and Webster Street in Malden, Massachusetts. A hedge surrounds the front yard.... The main entrance is a large wooden door with a circular window and is accessed by a set of stairs rising from a sidewalk running alongside Laurel Street. At the front door, there are four black mail boxes, two on each side of the door. [Unit D 3 ] is located on the second floor.... [Unit] D is' accessed by walking up the main staircase to the second floor landing. The door is on the right hand side. It is the only door on the right hand side of the second floor landing.

Bain’s name was not on any of the four mailboxes.

After entering the building, the agents “tried the keys in one door on the first floor and two doors on the second floor.” The keys fit the door to unit D. They entered unit D and conducted a “protective sweep to make sure no one else was inside.” The unit was empty. During the protective sweep, agents observed mail addressed to “131 Laurel Street, Apartment D, Malden, Massachusetts,” a parking ticket issued to Bain’s car on a chair in one- of the bedrooms, and a safe in- the same bedroom.

Armed with the information that the keys seized from Bain opened the main door to 131 Laurel Street and the door to unit D, the officers sought a warrant.to search unit D. In addition to the information summarized above, the affidavit used to obtain the warrant contained a series of statements, based on Connerney’s training and experience,^ establishing that it was reasonable to expect that Bain kept drugs, tools of the trade, cash, and records in the place where he resided.

Upon review of the affidavit, a federal magistrate judge issued a warrant to search unit D for a long list of items, including records relating to the purchase and sale of controlled substances, , cash derived from the sale of controlled substances, documents relating to the control of unit D, photographs of relevant property, and personal electronic devices. The subsequent search produced several key pieces of evidence against Bain. In one of the bedrooms, the police found a parking ticket for Bain’s car (the same one they saw during their warrantless entry into the unit), $300 in cash, and, in a trash can, used latex gloves and sandwich bags with the corners torn off. In a closet of that bedroom, they found several cards with Bain’s name on them: a Massachusetts driver’s license, a social security card, an identification card, an auto insurance card, a MassHealth card, and a AAA card. They also found a box in the closet, originally for size twelve Timberland shoes, that contained plastic bags, latex gloves, a digital scale, a bag containing 26.8 grams of a mixture of heroin and fentanyl, a razor blade, assorted pills, á handgun with an obliterated serial number, and a magazine containing bullets. In the closet of the other bedroom, police found a credit-card-making machine and boxes of blank cards, men’s clothing, a pair of Timberland boots, several boxes of sneakers (sizes eleven to twelve), and sneakers containing around $7000 in cash. Five of the $20 bills in the shoes came from the government funds that the cooperating witness had used in the March 21 controlled buy. At trial, the government submitted evidence that Bain wore size twelve shoes.

The district court denied Bain’s motion to suppress both the fact that the keys found in his possession opened the door to unit D and the items found during the search pursuant to the warrant. See United States v. Bain, 155 F.Supp.3d 107, 126 (D. Mass. 2015). In so doing, the court agreed with Bain that the turning of the key in the lock of unit D and the ensuing “protective sweep” were unreasonable searches. See id. at 120-23. Nevertheless, finding that the officers relied in good faith on precedent when turning the key, see id. at 121-22, the court took, account of the fact that the key fit the lock of unit. D when concluding that there was probable cause for a warrant to search unit D, see id. at 124. Bain challenges that ultimate ruling, and the government challenges the predicate ruling that testing the key in the lock was an unreasonable search. 4 The government also argues,. as a threshold matter, that Bain lacked sufficient connection to unit. D . to challenge the search.

B.

As We will explain, we find that Bain did have a sufficient connection with unit D to mount an -unfettered challenge to the search of that unit. In a matter of first impression in this circuit, we also find that the turning of the key in the lock of unit D was an unreasonable, warrantless search unsupported by any clear precedent, and that without the' information obtained by turning the key, there was no probable cause to issue a warrant to search unit D. Nevertheless, as we will also explain, because the officers were, entitled to rely in good faith on the warrant, the information secured in executing that warrant need not have been suppressed. 5

1.

We start by briefly, describing the tests used to determine whether a search has occurred within the meaning of the Fourth Amendment. Under the reasonable expectations test described in Justice Harlan’s concurring opinion in Katz v. United States, 389 U.S. 347, 88 S.Ct. 507, 19 L.Ed.2d 576 (1967), a search occurs whenever the government intrudes upon any place in which a person has a “reasonable expectation of privacy.” Id. at 360, 88 S.Ct. 507 (Harlan, J., concurring). There are two steps involved in applying this test. “First, we ask whether the individual, by his conduct, has exhibited an actual expectation of privacy; that is, whether he has shown that he ‘[sought] to preserve [something] as private.’” Bond v. United States, 529 U.S. 334, 338, 120 S.Ct. 1462, 146 L.Ed.2d 365 (2000) (alterations in original) (quoting Smith v. Maryland, 442 U.S. 735, 740, 99 S.Ct. 2577, 61 L.Ed.2d 220 (1979)). “Second, we inquire whether the individual’s expectation of privacy is one that society is prepared to recognize as reasonable.” Id.

The Supreme Court has also employed common-law trespass concepts to determine when a search has occurred. Under the common-law trespassory test described in Florida v. Jardines, 569 U.S. 1, 133 S.Ct. 1409, 185 L.Ed.2d 495 (2013), “[w]hen ‘the Government obtains information by physically intruding* on persons, houses, papers, or effects, ‘a “search” within the original meaning of the Fourth Amendment’ has ‘undoubtedly occurred.’ ” Id. at 1414 (quoting United States v. Jones, 565 U.S. 400, 404-05, 132 S.Ct. 945, 181 L.Ed.2d 911 (2012)); see also Grady v. North Carolina, — U.S. -, 135 S.Ct. 1368, 1370, 191 L.Ed.2d 459 (2015) (per curiam). This test supplements, rather than replaces, the Katz test. See Jardines, 133 S.Ct. at 1417 (“The Katz reasonable-expectations test ‘has been added to, not substituted for,’ the traditional property-based understanding of the Fourth Amendment, and so is unnecessary to consider when the government gains evidence by physically intruding on constitutionally protected areas.” (quoting Jones, 565 U.S. at 409, 132 S.Ct. 945)).

In Jardines itself, the Supreme Court employed the common-law trespas-sory test to determine that a physical intrusion into the “curtilage” of a home constituted a search under the Fourth Amendment even though no intrusion into the home had occurred. See id. at 1417-18. The curtilage is the area “immediately surrounding and associated with the home,” and it is “part of the home itself for Fourth Amendment purposes.” Id. at 1414 (quoting Oliver v. United States, 466 U.S. 170, 180, 104 S.Ct. 1735, 80 L.Ed.2d 214 (1984)). “This area around the home is ‘intimately linked to the home, both physically and psychologically,’ and is where ‘privacy expectations are most heightened.’ ” Id. at 1414-15 (quoting California v. Ciraolo, 476 U.S. 207, 213, 106 S.Ct. 1809, 90 L.Ed.2d 210 (1986)).

Under Jardines, a physical intrusion into a protected area that results in the acquisition of information only fails to constitute a search if that intrusion is permitted by a license. “[T]he knocker on the front door is treated as an invitation or license to attempt an entry, justifying ingress to the home by solicitors, hawkers and peddlers of all kinds.” Id. at 1415 (quoting Breard v. Alexandria, 341 U.S. 622, 626, 71 S.Ct. 920, 95 L.Ed. 1233 (1951)). “This implicit license typically permits the visitor to approach the home by the front path, knock promptly, wait briefly to be received, and then (absent invitation to linger longer) leave.” Id. The police may take advantage of this license to “approach a home and knock” without a warrant. Id. at 1416. However, “[t]he scope of a license—express or implied—is limited not only to a particular area but also to a specific purpose.” Id. The police behavior considered in Jardines—“introducing a trained police dog to explore the area around the home in hopes of discovering incriminating evidence,” id.—exceeded the implicit license. The Court provided other examples of behavior that would also exceed the license: “exploring the front path with a metal detector” or “marching [a] bloodhound into the garden before saying hello and asking permission.” Id

2.

The government contends that Bain lacks full Fourth Amendment rights in unit D. The parties refer to this issue as one of “standing” to assert Fourth Amendment rights. Although courts sometimes use this nomenclature, see United States v. Stokes, 829 F.3d 47, 51 (1st Cir. 2016), the Supreme Court has made clear that “definition of [Fourth Amendment] rights is more properly placed within the purview of substantive Fourth Amendment law than within that of standing,” Rakas v. Illinois, 439 U.S. 128, 140, 99 S.Ct. 421, 58 L.Ed.2d 387 (1978); see also Stokes, 829 F.3d at 51 n.7.

Bain was staying in unit D with his girlfriend, who rented the unit. The district court found that Bain “was, at the least, an overnight guest” in unit D. Bain, 155 F.Supp.3d at 115. Under Supreme Court precedent, Bain’s “status as an overnight guest is alone enough to show that he had an expectation of privacy in the home that society is prepared to recognize as reasonable.” Minnesota v. Olson, 495 U.S. 91, 96-97, 110 S.Ct. 1684, 109 L.Ed.2d 85 (1990). In short, Bain’s status as an overnight guest endowed him with Katz’s protection of a reasonable expectation of privacy in the unit.

The government nevertheless contends that Bain’s interest in the unit as an overnight guest, while sufficient to secure a reasonable expectation of privacy within the unit, falls short of the type of property ownership that would allow him to complain of a trespass within the unit or its curtilage. Hence, argues the government, an extension of Fourth Amendment interests derived from Jardines’s common-law trespassory test offers no protection to Bain as an overnight guest.

We reject this argument that a search defined in part by an invasion of property rights is a search only as to persons who could maintain a common law trespass claim. The property rights test applied in Jones and Jardines was foreshadowed by Justice Scalia’s concurring opinion in Minnesota v. Carter. 525 U.S. 83, 92-97, 119 S.Ct. 469, 142 L.Ed.2d 373 (1998) (Scalia, J., concurring). Justice Sca-lia stated that he considered Olson’s extension of Fourth Amendment rights to an overnight guest compatible with the property-based test because “it is plausible to regard a person’s overnight lodging as at least his ‘temporary’ residence.” Id. at 96-97, 119 S.Ct. 469. He explained that this conclusion is supported by both history and common understanding. See id. at 95-96, 119 S.Ct. 469 (stating that “[p]eople call a house ‘their’ home when legal title is in the bank, when they rent it, and even when they merely occupy it rent free—so long as they actually live there”). We agree with this reasoning. If a living unit is a person’s home under Olson, then the person’s Fourth Amendment protections are not diminished by the temporary nature of the person’s residence.

Therefore, to the extent that the key-turning is deemed a Fourth Amendment search because it constituted a trespassory invasion under Jones and Jardines—a subject that we will next discuss—we see no reason not to apply the amendment’s pro-teetions to an overnight guest’just as we would to a renter or owner.

3.

Having concluded that Bain has Fourth Amendment rights in unit D under both the reasonable-expectations test and the common-law trespassory test, we ask next whether a search occurred. Bain has argued that there was a search under both tests.

“At the [Fourth] Amendment’s ‘very core’ stands ‘the right of a man to retreat into his own home and there be free from unreasonable governmental intrusion.’” Jardines, 133 S.Ct. at 1414 (quoting Silverman v. United States, 365 U.S. 505, 511, 81 S.Ct. 679, 5 L.Ed.2d 734 (1961)), There is no reason to expect a different answer when the home is & rented condominium. See, e.g., Chapman v. United States, 365 U.S. 610, 615, 81 S.Ct. 776, 5 L.Ed.2d 828 (1961) (rented premises); Johnson v. United States, 333 U.S. 10, 17, 68 S.Ct. 367, 92 L.Ed. 436 (1948) (hotel room).

One might reasonably conclude that the inside of the front door lock is within the home itself because it is within the outer plane of the home’s structure.; Under Jardines, however, all we need decide is whether the inside of the front door lock is at least within the home’s curtilage. Under United States v. Dunn, 480 U.S. 294, 107 S.Ct. 1134, 94 L.Ed.2d 326 (1987), “the centrally, relevant consideration” in determining .the extent of a home’s curti-lage is “whether the area in question is so intimately tied to the home itself that it should be placed- .under the home’s ‘umbrella’ of Fourth Amendment protection.” Id. at 301, 107 S.Ct. 1134. To help answer this question, we consider four factors: (1) “the proximity of the area claimed to be curtilage to the home”; (2) “whether the area is included within an enclosure surrounding the home”; (3) “the nature of the uses to which the area is put”; and (4) “the steps taken by the resident to protect the area from observation by people passing by.” Id.

Applying the Dunn factors, 6 we conclude that the lock'on the door to unit D is within the unit’s curtilage even if it is not within the unit itself. 7 The first factor is strongly satisfied. Very few, if any, things are more proximate to the interior of a, home, than is a lock on the door to the home. Certainly, too, the interior of the lock, from which the crucial information was gathered, is within -or adjacent to the enclosure of the door’s outer face. The uses of the lock also strongly weigh in favor of finding.its penetration to be a search. The lock, after all, is used precisely to bar unwelcome entry and invasion of privacy. Finally, the very design of a lock hides its interior from examination. All in all, we have no difficulty finding that the inside of the lock on the door of a home “should be placed under the home’s ‘umbrella’ of Fourth Amendment protection.” Id.

Under Jardines, a physical intrusion into the curtilage to obtain information (here, putting the key in the lock to see if it fit) is a search unless it is within the “implicit license” which “typically permits the visitor to approach the home by the front path, knock promptly, wait briefly to be received, and then (absent invitation to linger longer) leave.” 133 S.Ct. at 1415. To be clear, assuming that the police were lawfully in the building, they could approach the door and knock without being deemed to haye conducted a search. But walking up to the door of a home and trying keys on the lock does not differ markedly from walking up with a trained police dog to sniff around the door. Paraphrasing Jardines: To find a visitor knocking on the door is routine (even if sometimes unwelcome); to find that same visitor trying a series of keys on the door’s lock “would inspire most of us to—well, call the police.” 133 S.Ct. at 1416. As in Jardines, “the background social norms that invite a visitor to the front door do not invite him there to conduct a search.” Id, In short, a search occurred.

c.

The government nevertheless argues that we are foreclosed from holding that trying the key on the door to unit D constituted a “search” by United States v. Lyons, 898 F.2d 210 (1st Cir. 1990), and United States v. Hawkins, 139 F.3d 29, 31 (1st Cir. 1998). We disagree. Lyons and Hawkins concerned the use of keys on storage container padlocks. Lyons, 898 F.2d at 212; Hawkins, 139 F.3d at 31. Storage container padlocks are “effects” under the Fourth Amendment. See Lyons, 898 F.2d at 219 (Woodlock, J., dissenting); cf. Oliver, 466 U.S. at 177 n.7, 104 S.Ct. 1735 (“The Framers would have understood the term ‘effects’ to be-limited to personal, rather than real, property.”); United States v. Jacobsen, 466 U.S. 109, 114, 104 S.Ct. 1652, 80 L.Ed.2d 85 (1984) (letters and sealed packages are effects); United States v. Place, 462 U.S. 696, 705-06, 103 S.Ct. 2637, 77 L.Ed.2d 110 (1983) (luggage is an effect); Jones, 565 U.S. at 404, 132 S.Ct. 945 (“It is beyond dispute that a vehicle is an ‘effect’ as that term is used in the [Fourth] Amendment.”). The Fourth Amendment protects effects markedly less than it protects houses. See Jardines, 133 S.Ct. at 1414 (“When it comes to the Fourth Amendment, the home is first among equals.”); Chambers v. Maroney, 399 U.S. 42, 52, 90 S.Ct. 1975, 26 L.Ed.2d 419 (1970) (“[F]oy the purposes of the Fourth Amendment there is a constitutional difference between houses and cars.”). Indeed, before Jardines the. Supreme Court held that dog-sniffs of certain effects are not searches under the Fourth Amendment. See Illinois v. Caballes, 543 U.S. 405, 409, 125 S.Ct. 834, 160 L.Ed.2d 842 (2005) (dog sniff of car); Place, 462 U.S. at 707, 103 S.Ct. 2637 (dog sniff of luggage). Jar-dines reached a different conclusion be^ cause it concerned a house rather than an effect. Justice Kagan’s concurrence in Jar-dines explained that Caballes did not control because the Court had “held, over and over again, that people’s expectations of privacy are much lower in their cars than in their homes.” 133 S.Ct. at 1419 n.1 (Kagan, J., concurring). Likewise, here, our statements in Lyons and Hawkins concerning the insertion of keys into padlocks on storage containers do not control whether testing a key on the lock to a home is a search.

The government also points to other circuit courts that have reached the conclusion that testing a key on a lock is not a search. Several of these cases involve the use of keys to identify the owners of cars, which are, as we have just explained, distinguishable. See United States v. $109,179 in U.S. Currency, 228 F.3d 1080, 1087-88 (9th Cir. 2000) (car door); United States v. DeBardeleben, 740 F.2d 440, 445 (6th Cir. 1984) (car door). The cases that involve testing keys on the doors to apartments are also either distinguishable or unconvincing. United States v. Salgado, 250 F.3d 438 (6th Cir. 2001), does involve testing a key in an apartment door. Salgado, however, was decided well before Jardines, and thus, at the time, plausibly rested on an observation that the lock on the apartment door (in an unlocked hallway) was “just as accessible to the public ... as an automobile lock.” Id. at 457. In finding that the Fourth Amendment protects the curtilage of a home from unlicensed searches even though it is readily accessible to the public, Jardines eliminated that basis for Salga-do’s holding. In United States v. Moses, 540 F.3d 263 (4th Cir. 2008), the Fourth Circuit held that “the discrete act of inserting the key into the lock and discovering whether or not it fit did not offend the Fourth Amendment.” Id. at 272. The opinion contains no reasoning or analysis. It merely cites Salgado, Lyons, $109,179 in U.S. Currency, and United States v. Concepcion, 942 F.2d 1170 (7th Cir. 1991). The first three cases we have discussed and distinguished above. Concepcion actually held that testing keys in an apartment door was a search, albeit one that was not unreasonable (a finding we will discuss in the next section of this opinion). See 942 F.2d at 1172-73. This case law provides no persuasive support for the government’s position, and we find ourselves comfortable in concluding that testing the key in the lock of unit D was a search.

4.

Having concluded that a search occurred when the police placed and turned a key in the lock of the door to unit D, we must determine whether that search was reasonable. The reasonableness of a search is a question of law, which we review de novo. See, e.g., United States v. Samboy, 433 F.3d 154, 158 (1st Cir. 2005).

The starting point for the reasonableness analysis is the “basic principle of Fourth Amendment law ... that searches and seizures inside a home without a warrant are presumptively unreasonable.” Kentucky v. King, 563 U.S. 452, 459, 131 S.Ct. 1849, 179 L.Ed.2d 865 (2011). And, as we have explained, Jardines treats unlicensed intrusions into the home’s curti-lage as intrusions into the home, see 133 S.Ct. at 1414, hence this presumption applies here even if we do not deem the door lock to be within the home itself. “But ... this presumption may be overcome in some circumstances because [t]he ultimate touchstone of the Fourth Amendment is reasonableness.” King, 563 U.S. at 459, 131 S.Ct. 1849 (alteration in original) (quoting Brigham City v. Stuart, 547 U.S. 398, 403, 126 S.Ct. 1943, 164 L.Ed.2d 650 (2006)). The main exceptions to this rule for house searches arise in cases of exigent circumstances, see id. at 460, 131 S.Ct. 1849 (listing the exceptions for “emergency aid,” for “hot pursuit” of a fleeing suspect, and for preventing “the imminent destruction of evidence”), and consent, see Georgia v. Randolph, 547 U.S. 103, 109, 126 S.Ct. 1515, 164 L.Ed.2d 208 (2006).

The exigent circumstances exception to the warrant requirement is a specific application of what the Supreme Court has described as a more general rule: “When faced with special law enforcement needs, diminished expectations of privacy, minimal intrusions, or the like, the Court has found that certain general, or individual, circumstances may render a warrantless search or seizure reasonable.” Illinois v. McArthur, 531 U.S. 326, 330, 121 S.Ct. 946, 148 L.Ed.2d 838 (2001); see also Maryland v. King, 569 U.S. 435, 133 S.Ct. 1958, 1969, 186 L.Ed.2d 1 (2013) (relying on same principle for search of person); United States v. Knights, 534 U.S. 112, 121, 122 S.Ct. 587, 151 L.Ed.2d 497 (2001) (similar for search of house); Maryland v. Buie, 494 U.S. 325, 331, 110 S.Ct. 1093, 108 L.Ed.2d 276 (1990) (similar). When a war-rantless search or seizure is not per se unreasonable, the court may “balance the privacy-related and law enforcement-related concerns to determine if the intrusion was reasonable.” McArthur, 531 U.S. at 331, 121 S.Ct. 946. As the name suggests, the balancing test requires assessing both the privacy interests and the law enforcement interests involved. See Arizona v. Hicks, 480 U.S. 321, 327, 107 S.Ct. 1149, 94 L.Ed.2d 347 (1987) (stating that a seizure is justified on less than probable cause when “the seizure is minimally intrusive and operational necessities render it the only practicable means of detecting certain types of crime”). The cases that apply this balancing test generally replace the warrant requirement with some other requirement, such as individualized suspicion at the level of reasonable suspicion or probable cause, see, e.g., Pennsylvania v. Labron, 518 U.S. 938, 940-41, 116 S.Ct. 2485, 135 L.Ed.2d 1031 (1996) (per curiam) (allowing warrantless search of automobile with probable cause); Place, 462 U.S. at 706, 103 S.Ct. 2637 (allowing temporary seizure of luggage based on reasonable suspicion); Terry v. Ohio, 392 U.S. 1, 27, 88 S.Ct. 1868, 20 L.Ed.2d 889 (1968) (allowing stop-and-frisk with reasonable suspicion), or with a limitation on the discretion of officers conducting the searches, see, e.g., King, 133 S.Ct. at 1969-70 (allowing mandatory buccal swabs of people arrested for serious crimes); Mich. Dep’t of State Police v. Sitz, 496 U.S. 444, 453-55, 110 S.Ct. 2481, 110 L.Ed.2d 412 (1990) (allowing mandatory stops at drunk driver checkpoint). When performing this balancing test, a court must look to the “totality of the circumstances.” Missouri v. McNeely, 569 U.S. 141, 133 S.Ct. 1552, 1559, 185 L.Ed.2d 696 (2013) (citing, inter alia, McArthur, 531 U.S. at 331, 121 S.Ct. 946). The government has the burden of proving that a warrantless search was nevertheless reasonable. See Vale v. Louisiana, 399 U.S. 30, 34, 90 S.Ct. 1969, 26 L.Ed.2d 409 (1970) (“[O]nly in ‘a few specifically established and well-delineated’ situations may a war-rantless search of a dwelling withstand constitutional scrutiny, even though the authorities have probable cause to conduct it. The burden rests on the State to show the existence of such an exceptional situation.” (citation omitted)).

The sum total of the government’s argument that the search of unit D’s lock was reasonable is (1) repeating that the intrusion involved in testing a key in a lock is “minor” or “minimal,” and (2) providing a string citation to a series of cases concluding that testing a key in a lock does not require a warrant or probable case. Two of the cases the government cites, Lyons and $109,179 in U.S. Currency, involve locks on effects rather than homes, as discussed above. One of the cases the government cites, Moses, contains no analysis, as we have also mentioned. We focus, therefore, on the government’s other two cited cases: United States v. Thompson, 842 F.3d 1002 (7th Cir. 2016), and Commonwealth v. Alvarez, 422 Mass. 198, 661 N.E.2d 1293 (1996). Both of these cases rely for their reasoning on an earlier Seventh Circuit case, United States v. Concepcion, 942 F.2d 1170 (7th Cir. 1991). See Thompson, 842 F.3d at 1008; Alvarez, 661 N.E.2d at 1302 & n.10. We do not think the reasoning of Concepcion adequately supports the government’s argument as to unit D. 8

In Concepcion, the Seventh Circuit held that, the use of a key in the lock of an apartment was a search, but that the search was reasonable without a warrant or probable cause. See 942 F.2d at 1172-73. The court reached the latter conclusion because

[w]here [the defendant] lived was something the agents could have ascertained in many other ways. They could have looked him up in the telephone book or conducted a computer search of drivers’ licenses. If they did hot find him (or if they found too many persons of the same name), they could have visited the landlord and asked who lived in apartment 1C. Instead of asking the landlord who lived there, they could have shown the landlord the key in their possession and asked the landlord to,compare it with the key issued to the tenant. So too the agents could have followed Concepcion around to learn his residence (as they did; the key just confirmed what they thought they knew). The information the agents obtained from putting the key in the lock thus was no secret.

Id. at 1173. The court contrasted the use of the key to the search found unreasonable in Hicks. There, a police officer who was lawfully present in an apartment to investigate gunfire noticed expensive stereo equipment including a turntable, moved the turntable to record its serial number, and then used the serial number to determine that the' turntable had been stolen. See 480 U.S. at 323, 107 S.Ct. 1149. The Supreme Court held that merely moving the turntable to read its serial number was an unreasonable search, even though the turntable itself was in plain view, because there was no probable cause to believe the turntable was stolen. See id. at 324-26, 326-27, 107 S.Ct. 1149. In Concepcion, the Seventh Circuit'concluded that trying the key oh the lock 'was a materially more minimal intrusion than moving the turntable because “[w]hat the officers learned from inverting the turntable in Hicks they could hot have come by in any other way,” while the agents in Concepcion “invaded less of [the defendant’s] interest in security of information' when they used the key to verify his address.” 942 F.2d at 1173. The court thus concluded that neither a warrant nor probable cause was required to try the key. Id.

We question the logic of justifying a search of this.type by reasoning that the information gathered by the search could have been easily obtained otherwise. After all, there are likely many pieces of information within a home that might be obtained from other sources without searching the home, It would seem, too, that the ease of obtaining information elsewhere undercuts law enforcement’s need to access the home more than it necessarily minimizes the nature of the intrusion into the home or its curtilage.

In any event, the government has made no argument and offered no evidence that it even'considered, much less pursued, other possible means of determining in which unit Bain resided. Nor does the government suggest that any exigencies in this case drove the need to turn the key in the lock of a home. No claim is made on appeal that evidence was being destroyed or that an imminent danger existed that the officers needed to enter unit P to address.

One might also say that the officers were merely trying to identify unit D as Bain’s residence, rather than searching unit D. Cf. DeBardeleben, 740 F.2d at 445. Of course, one could equally say that the officers in Hicks were merely trying to identify the turntable. The key point is that the officers intruded without license or warrant into the curtilage of Bain’s “home” solely to gather information to be used in building a criminal case against him. In short, we see no reason to conclude that the law enforcement-related concerns sufficiently outweighed the privacy-related concerns to render this search reasonable. 9

5.

Although we have concluded that there was an unreasonable search in violation of the Fourth Amendment, that does not necessarily mean that the evidence seized pursuant to the warrant must be suppressed. A number of exceptions to the éxclusion-ary rule exist, and the government has argued two of them; good-faith reliance on clear precedent and good-faith reliance on a warrant. 10

The government argues that the officer who turned the key relied in good faith on precedent and, .therefore, the exclusionary rule should not apply to the information obtained either from that war-rantless search or from the search warrant obtained with affidavits containing that information. • This argument relies on. Davis v. United States, 564 U.S. 229, 131 S.Ct. 2419, 180 L.Ed.2d 285 (2011), in which the Supreme Court held that evidence obtained from a warrantless search performed in good-faith reliance on binding precedent should not be subject to the. exclusionary rule. See id. at 235, 240-41, 131 S.Ct. 2419. As applied here, this argument also; presumes that- if information obtained in reliance on clear precedent should not be suppressed, it also should not be excised from a warrant affidavit in deciding whether there was probable cause to issue the warrant. Cf. United States v. Dessesaure, 429 F.3d 359, 367 (1st Cir. 2005).

The district court accepted these arguments. It held that trying the key on the lock of unit D was a search and a Fourth Amendment violation, but that the police “reasonably relied” on this court’s earlier opinions in Lyons and Hawkins. See Bain, 155 F.Supp.3d at 121-22. The district court then concluded, implicitly, that information obtained in good-faith reliance on precedent need not be excluded from a warrant affidavit when determining whether the independent source doctrine applies under Dessesaure. See id. at 124.

This court has clarified that “the [Davis] exception is available only where the police rely on precedent that is ‘clear and well-settled.’” United States v. Sparks, 711 F.3d 58, 64 (1st Cir. 2013) (quoting United States v. Davis, 598 F.3d 1259, 1266 (11th Cir. 2010)). “[T]his emphasis on the clear application of the precedent to the case at hand is consistent with Davis’s focus on deterrence; where judicial precedent does not clearly authorize a particular practice, suppression has deterrent value because it creates an ‘incentive to err on the side of constitutional behavior.’ ” Id. (quoting Davis, 598 F.3d at 1266-67); see also United States v. Whitaker, 820 F.3d 849, 854-55 (7th Cir. 2016) (requiring an on-point holding); United States v. Burston, 806 F.3d 1123, 1129 (8th Cir. 2015) (requiring a holding in a similar factual context).

We do not think that Davis covers the police conduct here. As discussed above, the facts of Lyons were quite different from the facts here and were different in ways that a reasonable person would suspect might be legally significant. That suspicion would have been heightened by Jones and Jardines, both of which were decided before the search in this case occurred in 2014. In light of Jardines, it could not have been “clear and well-settled” that Lyons would apply to testing keys on the locks of houses. Indeed, Lyons relied on authority that clearly did not apply to houses in light of Jardines: Lyons cited Place to support its conclusion that testing the key on the padlock was not a search. See Lyons, 898 F.2d at 213 (citing Place, 462 U.S. at 707, 103 S.Ct. 2637). Place held that using a drug-detecting dog to sniff luggage was not a Fourth Amendment search. See 462 U.S. at 707, 103 S.Ct. 2637. Jardines reached a holding directly to the contrary with respect to houses. Several of the opinions in Jardines highlighted this distinction. See Jardines, 133 S.Ct. at 1419 n.1 (Kagan, J., concurring); id. at 1424 (Alito, J., dissenting). In light of these distinctions, we cannot agree that police were acting in accordance with precedent that was “clear and well-settled.” Sparks, 711 F.3d at 64. 11

The government argues in the alternative that no suppression should result because the officers who searched unit D relied in good faith on the magistrate’s issuance of a warrant, even though the affidavit contains information obtained in violation of the Fourth Amendment. In support of this argument, the government points to United States v. Leon, 468 U.S. 897, 104 S.Ct. 3405, 82 L.Ed.2d 677 (1984). Under Leon, evidence obtained from a search conducted “in objectively reasonable reliance on a subsequently invalidated search warrant” need not always be excluded. See id. at 922, 104 S.Ct. 3405. Although the existence of a warrant issued by a magistrate will usually establish this form of good faith, “in some circumstances the officer will have no reasonable grounds for believing that the warrant was properly issued.” Id. at 922-23, 104 S.Ct. 3405 (footnote omitted). The Leon court provided examples of four such circumstances: (1) “if the magistrate or judge in issuing a warrant was misled by information in an affidavit that the affiant knew was false or would have known was false except for his reckless disregard of the truth,” id. at 923, 104 S.Ct. 3405 (citing Franks v. Delaware, 438 U.S. 154, 98 S.Ct. 2674, 57 L.Ed.2d 667 (1978)); (2) “where the issuing magistrate wholly abandoned his judicial role,” id.; (3) when an affidavit is “so lacking in indicia of probable cause as to render official belief in its existence entirely unreasonable,” id. (quoting Brown v. Illinois, 422 U.S. 590, 610-611, 95 S.Ct. 2254, 45 L.Ed.2d 416 (1975) (Powell, J., concurring in part)); and (4) when, “depending on the circumstances of the particular case, a warrant [is] so facially deficient—i.e., in failing to particularize the place to be searched or the things to be seized—that the executing officers cannot reasonably presume it to be valid.” Id. We review the application of the good-faith exception de novo. See United States v. Baez, 744 F.3d 30, 33 (1st Cir. 2014).

Here, we have a circumstance not expressly addressed in Leon: the warrant affidavit forthrightly discloses facts that establish probable cause, but one of the facts essential to establishing probable cause (the result of the key turn) was obtained as a result of an unconstitutional search. We encountered a very similar circumstance in United States v. Diehl, 276 F.3d 32 (1st Cir. 2002). In Diehl, the defendant sought suppression of evidence “seized pursuant to a facially valid warrant.” Id. at 34. As in this case, the warrant affidavit contained a report that an officer had previously approached the searched location and this report “was necessary to establish the probable cause justifying issuance of the warrant.” Id. at 35, 41-42. Anticipating Jardines, this court concluded that the officer’s earlier visit to the location was an unconstitutional search because it involved a warrantless trespass on the curtilage of the residence. See id. at 38, 41. Nevertheless, we concluded that Leon’s good faith exception applied. In so ruling, we focused on the accuracy and completeness of the manner in which the information supporting the warrant was conveyed to the magistrate issuing the warrant. Placing the burden on the government, id. at 42, we asked whether the affiant’s recitation of the facts was infected “with an intentional misrepresentation, or one made with reckless disregard of the truth,” so as to mislead the magistrate. Id. We asked as well whether, by omission or error, the description “[took] away from the issuing court the ability to decide” the curtilage issue for itself. Id. at 42-43. We also asked whether enough information was given to the issuing judge to determine whether the officer who invaded the curtilage acted “in such bad faith as to preclude a warrant.” Id. at 43. Finally, we noted the possibility that snow cover may have misled the officers as to the contours of the curtilage, which would further negate any inference of bad faith. Id. All in, we found the case to present “ ‘a penum-bral zone’ within which an inadvertent mistake would not call for exclusion.” Id. (quoting Leon, 468 U.S. at 925 n.26, 104 S.Ct. 3405).

Diehl’s application of the Leon good-faith exception finds company in the majority of circuits- that have considered the question posed by reliance on a warrant that is itself tainted by the results of. an unconstitutional search. See, e.g., Hopkins, 824 F.3d at 733; United States v. Ganias, 824 F.3d 199, 222-23 (2d Cir. 2016) (en banc); United States v. Massi, 761 F.3d 512, 528 (5th Cir. 2014); United States v. McClain, 444 F.3d 556, 565-66 (6th Cir. 2005). At least two circuits have disagreed with this majority view, see, e.g., United States v. McGough, 412 F.3d 1232, 1239-40 (11th Cir. 2005); United States v. Wanless, 882 F.2d 1459, 1466-67 (9th Cir. 1989), as have a few commentators, see, e.g., 1 Wayne R. LaFave, Search & Seizure: A Treatise, on the Fourth Amendment § 1.3(f) (5th ed. 2016) (stating that “there is good reason to doubt” whether the plurality rule is correct); Craig M. Bradley, The “Good Faith Exception” Cases: Reasonable Exercise in Futility, 60 Ind. L.J. 287, 302 (1985) (“When the magistrate issued the warrant, he did not endorse past activity; he only authorized future activity.... [T]he function of the magistrate is to determine ‘whether a particular affidavit establishes probable cause,’ not whether the methods used to obtain the information in that affidavit were legal." (quoting Leon, 468 U.S. at 914, 104 S.Ct. 3405)). Diehl would seem to suggest that the minority position takes too cramped a view of what magistrates do, and accords too much relevance to a distinction that may have no bearing on the presence or absence of good faith. Diehl, 276 F.3d at 42-43. Be that as it may, this case presents no reason to deviate from Diehl’s interpretation of Leon. Under Diehl, good faith reliance on a warrant procured and issued in good faith saves the fruits of a warranted search from suppression.

So, we turn to the question of good faith. Unlike in Diehl, the invasion of the curtilage in this case could not be said to be the result of the officer’s misapprehension of the facts. Here, any misapprehension was purely a misapprehension of the law. Diehl offers no direct guidance on how to define the point at which such a misapprehension equates with the bad faith that would negate reliance on the warrant. Pointing to United States v. Hopkins, 824 F.3d 726 (8th Cir.), cert. denied, — U.S. -, 137 S.Ct. 522, 196 L.Ed.2d 425 (2016), the government' says that we should find the Leon good faith exception applicable because the key-turn was “close enough to the line of validity to make the officers’ belief in' the validity of the warrant objectively reasonable.” Id. at 733. Bain declines to argue that we should not employ Hopkins’s formulation. Rather, he argues that the officers’ conduct here fails to qualify under that formulation. Hopkins’s focus on objective reasonableness seems to, align with Leon’s repeated references to reasonableness. This alignment, coupled with Bain’s failure to contest the application of Hopkins, leads us to employ Hopkins’s formulation, albeit by assuming rather than deciding that such a formulation is the proper one for measuring the officer’s good faith.

Applying this formulation, we conclude that the police could rely in, good faith on the search warrant in this case. As we have explained, our decisions in Lyons and Hawkins did not clearly classify the turning of a key in an apartment lock as being a reasonable search. Warrants, though, make a difference. Once the magistrate issued a warrant, the relevant question was no longer whether clear precedent blessed the search upon which the warrant was based in part. Rather, the question became whether precedent pointed enough in that direction to allow an objectively reasonable officer informed .about the law to conclude (erroneously, as. we have now explained) that he could turn a key in the lock of unit D on the basis of a reasonable suspicion short of probable cause.

We think that reasonable officers informed about the law (prior to the issuance of this opinion) could have so concluded. Indeed, the Massachusetts Supreme Judicial Court had so concluded, holding that only reasonable suspicion was required for just such a search. See Alvarez, 661 N.E.2d at 1302. So, too, as we have noted, did the Seventh Circuit in Concepcion, and it did so on grounds not directly rejected in Jardines. See Concepcion, 942 F.2d at 1172-73. Given the facts known to the officers at the time they tried the keys, it was reasonable to suspect that turning the key on the lock to unit D would lead to evidence of Bain’s drug dealing. There was good reason to believe Bain was -residing, at least temporarily, in one of the apartments accessible through the front door of 131 Laurel Street: He had been seen there previously both during the day and late at night, and he walked out that front door right before his arrest.' His keys, in turn, did not work in the doors of two other units, leaving a fifty percent chance that they would fit unit D. 12 As we will discuss in the next section, there was a nexus between Bain’s drug dealing and the location where he was residing. Given the presence of reasonable suspicion, and given the state of the law prior to today’s decision holding the key turning to constitute an unlawful search, checking the keys on the door to unit'D was sufficiently close to the line of validity that the police could rely in good faith on'the search-warrant.

Finally, Bain raises an alternative argument for suppressing the fruits of the warranted search. He contends that the warrant was defective because the affidavits provided no probable cause to believe that evidence of his suspected crime would be found in unit D, even without excising the fact that a key in his possession opened unit D. The question posed by this argument is whether the warrant affidavits were “so lacking in indicia of probable cause as to render official belief in its existence entirely unreasonable.” Leon, 468 U.S. at 923, 104 S.Ct. 3405.

Bain does not challenge that the police had' probable cause to believe that he committed a crime; instead, he argues that the search warrant affidavits did not establish an adequate nexus 'between that crime and unit D. “When it comes to nexus, common sense says that a connection with the search site can be'deduced ‘from the type of crime, the nature of the items sought,’ plus ‘normal inferences as to where a criminal would hide’ evidence of his crime.” United States v. Rivera, 825 F.3d 59, 63 (1st Cir. 2016) (quoting United States v. Feliz, 182 F.3d 82, 88 (1st Cir. 1999)). This court has, “with a regularity bordering on the echolalic, endorsed the concept that a law enforcement officer’s training and experience may yield insights that support a probable cause determination.” United States v. Floyd, 740 F.3d 22, 35 (1st Cir. 2014). We also" have a line of precedent that addresses wheri police have probable cause to search the homes of people known to be selling drugs. See United States v. Barnes, 492 F.3d 33, 37 (1st Cir. 2007); United States v. Ribeiro, 397 F.3d 43, 49 (1st Cir. 2005); Feliz, 182 F.3d at 87-88. We have’ expressed skepticism that probable' cause can be established by the combination of the fact that a defendant sells drugs and general information from police officers that drug dealers tend to store evidence in their homes. See Ribeiro, 397 F.3d at 50-51 (citing United States v. Schultz, 14 F.3d 1093, 1097 (6th Cir. 1994)); Feliz, 182 F.3d at 87-88. However, the addition of specific facts connecting the drug dealing to the home can establish a nexus. See Ribeiro, 397 F.3d at 51.

The warrant affidavits established that Bain had access to unit D. True, they did so by relying on the unlawfully obtained information establishing that the key fit the lock on unit D. But as we have already explained, that defect did not undermine the officers’ ability to rely on the warrant. So, the only question was whether Bain was staying there in the sense that one would expect to find his possessions there. We think the affidavits create a fairly strong inference that he was. Officer Con-nerney stated that when Bain lived with his brother in Waltham, Connerney regularly observed Bain’s car parked outside. After Bain was arrested for assaulting his brother, he told his probation officer that he would live at his mother’s residence in Arlington. Connerney stated that he had never seen Bain at that residence and had never seen Bain’s car parked outside that residence. Connerney also stated that since the controlled purchase on March 21, officers had seen Bain’s car parked on Webster Street in Malden, near 131 Laurel Street, on several occasions. On two of those occasions, March 28 and 29, officers could have reasonably inferred that Bain stayed in the area overnight.

Bain also argues that even if there was probable cause to believe he was living in unit D, the affidavits do not establish probable cause to believe that evidence of his heroin dealing would be found there. The affidavits raise the inference, though, that Bain’s practice was to deliver the product near a location where he resided, first near his brother’s apartment in Waltham, and then near 131 Laurel Street after he moved. And at the moment that he was arrested exiting the home, he was apparently carrying (or a reasonable officer could infer he was carrying) a relatively small amount of heroin. Neither his person nor his car contained any of the items that a repeat drug seller very often has (a source stash, cash, and records). 13 Neither his altercation-induced exit from his brother’s apartment nor any other information in the warrant affidavits even hinted at any other location that might contain the expected accoutrements of an experienced dealer. All in all, and whether or not this added up to probable cause of a nexus to unit D, it was enough so that we cannot say that the nexus-related evidence upon which the warrant rested was “so lacking in indicia of probable cause as to render official belief in its existence entirely unreasonable.” Leon, 468 U.S. at 923, 104 S.Ct. 3405.

d.

Finally, we consider whether the district court erred in failing to suppress testimony at trial from two police officers describing how they used Bain’s keys to open the door to unit D immediately after his arrest. We have concluded that the police could not have relied in good faith on precedent to conduct this warrantless search in the first instance, but that they could rely in good faith on the warrant when conducting the later warranted search. In short, the direct results of the warrantless key-turn search were inadmissible, but that infirmity did not taint the results of the subsequent warranted search. These conclusions have the following consequence: Unless the warranted search also revealed that the keys in Bain’s possession at the time of his arrest fit the lock on the door to unit D, it was eiTor to admit this testimony.

The government has not pointed us to any testimony establishing that the police retested the keys on the lock after the warrant issued. We have not found any. 14 Therefore, the district court should have suppressed the officers’ testimony about using Bain’s keys to open the door to unit D immediately after his arrest.

Nevertheless, we conclude that admitting this testimony was harmless beyond a reasonable doubt. See United States v. Rose, 802 F.3d 114, 124 (1st Cir. 2015) (affirming denial of motion to suppress because, even if defendant were correct that evidence should have been suppressed because there was no independent source, the “government [could] prove beyond a reasonable doubt that the [putative] error complained of did not contribute to the verdict obtained.” (second alteration in original) (quoting United States v. Green, 698 F.3d 48, 53-54 (1st Cir. 2012))). As in Rose, although the government referred to the improperly obtained evidence in both its opening and closing arguments, “the remaining evidence was so overwhelming that, even if this evidence should have been excluded, its inclusion did not affect the verdicts.” Id. As we have summarized above, the warranted search uncovered overwhelming evidence that Bain resided, at least temporarily, in unit D. His government-issued identification cards were in one of the closets, along with a health insurance card, auto insurance card, and AAA card, all in his name. The other closet contained a large quantity of men’s clothing and shoes in Bain’s size, including a sneaker containing $20 bills that the cooperating witness had used to purchase heroin from Bain. A parking ticket for Bain’s car was found on a chair. We have no doubt that the jury would have concluded that Bain possessed the drugs and gun in unit D even if the court had excluded the testimony about his keys.

II.

Bain argues that even if the suppression motion properly failed, the district court still should not have admitted at trial evidence that there was a credit-card-making machine in one of the closets of unit D. We review the judge’s evidentiary ruling admitting this evidence for abuse of discretion. See United States v. Gemma, 818 F.3d 23, 35 (1st Cir. 2016); United States v. Varoudakis, 233 F.3d 113, 118 (1st Cir. 2000).

As described in section I.A, supra, in the closet of one of the bedrooms in unit D, police found a credit-card-making machine inside an opaque trash bag along with boxes of blank cards. A government witness testified that when Bain was arrested, the police found four credit cards on his person. Another witness testified that those credit cards were fake and that the machine found in the closet could have been used to make them. In the same closet as the credit-card-making machine, police found men’s sneakers, which contained around $7,000 in cash. Five of the $20 bills in the shoes were government funds that the cooperating witness had used in the March 21 controlled buy.

Bain challenged the admission of the credit-card-making machine in a motion in limine. The district court concluded that “[t]he Government may present evidence regarding the credit card making equipment for the purpose of. connecting .Defendant to the apartment in question and establishing that Defendant had control over the area in which .the items were found.” At trial the district court gave two limiting instructions on this evidence. The first was that the jury should not use the evidence to decide whether “other wrongful acts did or did not occur.” The second was that the jury should not “consider the evidence as proof that the Defendant [ha]s a bad character or any propensity to commit crime.”

A proposal by the government to introduce evidence of a defendant’s other bad acts is subject to a two-part test. See United States v. Hicks, 575 F.3d 130, 142 (1st Cir. 2009). “First, a court must ask whether the proffered evidence has a ‘special’ relevance, i.e., a non-propensity relevance.” Id. Under Rule 404(b), “[evidence of a crime ... is not admissible to prove a person’s character in order to show that 'on a particular occasion the person acted in accordance with the character.” Fed. R. Evid. 404(b)(1). But such evidence “may be admissible for another purpose, such as proving motive, opportunity, intent, preparation, plan, knowledge, identity, absence of mistake, or lack of accident.”' Fed. R. Evid. 404(b)(2). If other bad acts evidence has special relevance under Rule 404(b), the court must consider whether the evidence should nevertheless be excluded under Rule 403. See Hicks, 575 F.3d at 142. “The court may exclude relevant evidence if its probative value is substantially outweighed by a danger of ... unfair prejudice.” Fed. R. Evid. 403.

Bain first argues that the machine had no special relevance under Rule 404(b). Conceding that “evidence tending to tie an individual to a certain location might, in some instances, have special relevance,” he argues, first, that the testimony about the capabilities of the machine did not match up with the nature of the fake cards in his pocket. He also argues that the machine did not connect him to unit D because there was no evidence that the cards in his pocket were uniquely connected to the machine, that he had made the cards himself, or that he knew they were fake.

Bain’s first argument relies on a factual predicate that one of the government’s witnesses contradicted. A secret service agent called by the government testified that the machine could have been used to make the fake credit cards in Bain’s pocket. Bain argues, essentially, that this testimony was contradicted by the witness’s prior statement that the machine was a “card embosser” combined with the (claimed) fact that the cards were not actually embossed. As is often the case, . this type of argument, which highlights a contradiction in a witness’s testimony and insists that one statement should be believed over another, is an argument about probative value. A jury could believe the agent’s statement that the machine could make the unembossed cards in Bain’s pocket notwithstanding the agent’s reference to the machine as a card embosser. 15

Bain’s latter arguments also attack probative value rather than special relevance. See, e.g., United States v. Gentles, 619 F.3d 75, 87 (1st Cir. 2010). The presence of the machine in the unit and the fake credit cards in Bain’s pocket, combined with the testimony that tire machine could make those cards, made it more'likely that he had been in the unit, that he stored his possessions in the unit, that he stored his possessions in that closet in particular, and thus that he controlled other items in that closet, including the sneaker with the money in it. This chain of inferences establishes special relevance. See Fed. R. Evid. 401 (defining “relevant” evidence); United States v. Doe, 741 F.3d 217, 229 (1st Cir. 2013) (stating that evidence has special relevance if “it is relevant for any purpose apart from showing propensity to commit a crime”). The existence of other possible chains of inference does not undermine special relevance; it goes to probative weight.

Bain next ar'gués that the machine’s probative value was substantially outweighed by the risk of unfair prejudice, so it should have been excluded under Rule 403; He posits that the probative value of the credit-card-making machine' was slight because it was cumulative of other evidence showing that he lived at the unit (e.g., observations of him leaving the building, the presence of his identification cards in the unit, the presence of money used in one of the controlled buys in the unit). This small probative value, he argues, was substantially outweighed by a risk the jury would infer he was engaged in credit-card fraud and, therefore, had a bad character.

When assessing the probative value of evidence under Rule 403, a court must consider both whether the evidence was offered to prove an issue that was in genuine dispute, and whether the eviden-tiary point could have been made with other evidence that did not. present a risk of unfair prejudice. See United States v. Ford, 839 F.3d 94, 109-10 (1st Cir. 2016); Varoudakis, 233 F.3d at 122. The risk of prejudice from admitting a piece of evidence may be “cabined” by a limiting instruction. See United States v. Pelletier, 666 F.3d 1, 6 (1st Cir. 2011).

The Rule 403 issue is a close one. Bain’s defense at trial was that: (1) the recordings of the controlled buys did not establish definitively that he sold drugs to the cooperating witness; (2) the cooperating witness was unreliable; and (3) the items in the unit were not his. He advanced these defenses both in opening and in closing. The government did have abundant evidence that Bain was connected to the unit, and the identification cards were far more probative of Bain’s control over the drugs, gun, and ammunition found in one of the closets than the credit-card-making machine. But Bain also denied that he had engaged in the controlled buys, making it crucial to the government’s case to establish his control over the money in the sneaker in the closet of the other bedroom. The credit-card-making machine was probative on that point. In light of this probative value, the issues disputed at trial, and the limiting instructions, we cannot conclude that the district court abused its discretion in admitting the credit-card-making machine. See United States v. Smith, 292 F.3d 90, 99, 100-01 (1st Cir. 2002) (noting that “[w]e usually defer to the district court’s balancing under Rule 403 of probative value against unfair prejudice” and factoring the district court’s limiting instruction into the Rule 403 analysis).

III.

Finally, we consider the sentencing issue. Bain argues that he did not have three prior convictions for “serious drug offenses,” and therefore did not qualify for the ACCA’s fifteen-year mandatory minimum sentence. 16 He did not preserve this argument below, so we review only for plain error. 17

The ACCA provides:

In the case of a person who violates section 922(g) of this title and has three previous convictions by any court ... for ... a serious drug offense ... committed on occasions different from one another, such person shall be ... imprisoned not less than fifteen years....

18 U.S.C. § 924(e)(1). A “serious drug offense” under the ACCA includes, in relevant part,

an offense under State law, involving manufacturing, distributing, or possessing with intent to manufacture or distribute, a controlled substance (as defined in section 102 of the Controlled Substances Act (21 U.S.C. 802)), for which a maximum term of imprisonment of ten years or more is prescribed by law[.]

Id. § 924(e)(2)(A)(ii).

The probation officer who prepared Bain’s Presentence Investigation Report (PSR) concluded that Bain had three prior convictions for serious drug offenses: a 2002 conviction for possession of cocaine with intent to distribute, a 2006 conviction for possession of crack cocaine with intent to distribute, and a 2006 conviction for trafficking twenty-eight to one hundred grams of cocaine, all under Massachusetts law. Bain did not object to this conclusion, and the district court accepted it. On appeal, Bain belatedly argues that his 2006 Massachusetts conviction for trafficking cocaine, see Mass. Gen. Laws ch. 94C, § 32E(b), does not fall under the definition of a “serious drug offense.”

Determining whether Bain’s prior conviction falls into this definition requires looking at the Massachusetts statute under which he was convicted and the relevant state precedent. The statute reads:

Any person who trafficks in a controlled substance [as defined to include cocaine] by knowingly or intentionally manufacturing, distributing or dispensing or possessing with intent to manufacture, distribute or dispense or by bringing into the commonwealth a net weight of 18 grams or more of a controlled substance as so defined, or a net weight of 18 grams or more of any mixture containing a controlled substance as so defined shall [be punished by a term of imprisonment that varies depending on weight].

Id. The parties’ dispute focuses on the form of trafficking committed by “bringing into the commonwealth a net weight of 18 grams or more of [cocaine], or a net weight of 18 grams or more of any mixture containing [cocaine].” Id. Bain argues that this form of the offense is not a serious drug offense. He does not dispute that the other forms of trafficking are serious drug offenses.

We need not and do not determine whether Bain is right that the “bringing into the commonwealth” form of the offense is not a serious drug offense. Instead, we assume that he is right 18 but conclude that he still cannot satisfy the plain error standard because he cannot establish that the Massachusetts trafficking statute is clearly “indivisible.” We pause for a moment to explain what that means.

When determining whether a prior conviction qualifies as a predicate offense under the ACCA, we do not look at the specific facts of the defendant’s prior conviction. Instead, we use a categorical approach, where we classify crimes as ACCA predicates based on their legal definitions, rather than the facts of the defendant’s particular conviction. See Mathis v. United States, — U.S. -, 136 S.Ct. 2243, 2248, 2251-52, 195 L.Ed.2d 604 (2016); Descamps v. United States, 570 U.S. 254, 133 S.Ct. 2276, 2283, 186 L.Ed.2d 438 (2013). The categorical approach is imposed in part by the language of the ACCA, see Johnson v. United States, — U.S. -, 135 S.Ct. 2551, 2562, 192 L.Ed.2d 569 (2015); Shepard v. United States, 544 U.S. 13, 19, 125 S.Ct. 1254, 161 L.Ed.2d 205 (2005); Taylor v. United States, 495 U.S. 575, 600, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990), and in part by the Sixth Amendment concerns that would arise if the imposition of the ACCA’s mandatory minimum sentence were based on the actual facts underlying prior convictions as found by the sentencing judge, see Shepard, 544 U.S. at 24, 125 S.Ct. 1254; Taylor, 495 U.S. at 601, 110 S.Ct. 2143; United States v. Faust, 853 F.3d 39, 50 (1st Cir. 2017) (citing Mathis, 136 S.Ct. at 2252, and Descamps, 133 S.Ct. at 2288).

The categorical approach proceeds in different ways for different portions of the ACCA. To determine whether a conviction for a crime falls within the “force clause” or the “enumerated offense clause” of the violent felony definition, see United States v. Starks, 861 F.3d 306, 314 (1st Cir. 2017) (defining these terms), courts ask whether there is any (realistic) way of committing the crime that does not satisfy the force clause or the elements of the generic version of the enumerated offense. See, e.g., Mathis, 136 S.Ct. at 2248 (enumerated offense clause); Faust, 853 F.3d at 51 (force clause). To determine whether a conviction for a crime falls within the now-invalidated “residual clause” of the violent felony definition, see Starks, 861 F.3d at 314, courts asked whether the “ordinary case” of a conviction for that crime created a risk of physical injury that exceeded a difficult-to-specify threshold. See, e.g., Johnson, 135 S.Ct. at 2557-59.

In this circuit, when analyzing whether a prior conviction is a serious drug offense, we have followed the approach used with the force clause and the enumerated offenses clause and asked whether (as determined from the crime’s definition and state cases implementing that definition) every realistically possible way of committing the offense satisfies the definition of a serious drug offense. See United States v. Mulkern, 854 F.3d 87, 96-97 (1st Cir. 2017); United States v. Whindleton, 797 F.3d 105, 109 (1st Cir. 2015) (“Since Whindleton’s record of conviction does not specify on what theory he was convicted, we must ensure that any form of the conviction would qualify as a ‘serious drug offense’ under the ACCA.”), cert. dismissed, — U.S. -, 137 S.Ct. 23, 195 L.Ed.2d 896, cert. denied, — U.S. -, 137 S.Ct. 179, 196 L.Ed.2d 147 (2016). This approach makes good sense, given that the difficulties involved in applying the ordinary case approach contributed to the conclusion that the residual clause was unconstitutionally vague. See Johnson, 135 S.Ct. at 2557-58.

Not all criminal statutes define only a single crime. “Some statutes ... have a more complicated (sometimes called ‘divisible’) structure.... A single statute may list elements in the alternative, and thereby define multiple crimes.” Mathis, 136 S.Ct. at 2249. “A sentencing court thus requires a way of figuring out which of the alternative elements listed .., was integral to the defendant’s conviction....” Id. “To address that need, th[e Supreme] Court approved the ‘modified categorical approach’ for use with statutes having multiple alternative elements.” Id. “Under that approach, a sentencing court looks to a limited class of documents (for example, the indictment, jury- instructions, or plea agreement and colloquy) to determine what crime, with what elements, a defendant was convicted of.” Id. (citing, inter alia, Shepard, 544 U.S. at 26, 125 S.Ct. 1254), We call this limited class of documents “Shepard documents.”

Not all'crimes that can be committed in multiple'different ways are' divisible into multiple crimes with different elements. There is “a different kind of alternatively phrased law: not one that lists multiple elements disjunctively, but instead one that enumerates various factual means of committing a single element.” Id. In order'to determine whether a crime that may be committed in multiple different ways is divisible, We must be able to distinguish between crimes that have alternative elements and crimes that have a single set of elements that may be satisfied by different means. There are a number of different ways of distinguishing elements from means, including looking at jury unanimity requirements, relevant model jury instructions, certain statutory provisions, and indictments. See Starks, 861 F.3d at 316.

Finally, “if state law fails to provide clear answers” about what are elements and áre mean's, “federal judges have another place to look: the record of a prior conviction "itself,” Mathis, 136 S.Ct. at 2256, that is, the Shepard documents. If neither state law nor the Shepard docur ments “speak[s] plainly” about whether a crime is divisible, a sentencing court must assume that it is not. See id. at 2257.

This case involves-another layer of complexity: the plain error standard. Generally, it is the government’s burden to prove that a defendant has three predicate convictions under the ACCA, See Mulkern, 854 F.3d at 90. Thus, when a statute is divisible and some forms of the offense are ACCA predicates and some forms are not, the government bears the burden of proving that the defendant was convicted of a form that is an ACCA predicate. See id. But that burden shifts on plain error review.-When a defendant fails to preserve an objection to the government’s contention that a prior conviction for a divisible offense was for the qualifying form of that offense, the defendant can only win .on appeal by proving, that the conviction was not for the qualifying form of the offense. See United States v. Serrano-Mercado, 784 F.3d 838, 846-49 (1st Cir. 2015), cert. denied, — U.S. -, 137 S.Ct. 812, 196 L.Ed.2d 599 (2017). 19

This case presents a. related issue of first impression: What do we do on plain error review when state law lacks clarity on the question of divisibility and the record lacks the Shepard documents to which we might otherwise refer in an attempt to resolve that ambiguity? Under Mathis, when state law is unclear about the elements of a crime, a sentencing court may consult the Shepard documents to determine whether the crime is divisible. Normally, the government must introduce these documents and prove that the crime is divisible in' order to carry its burden of proving that the defendant was convicted of a form of an offense that qualifies as an ACCA predicate. See United States v. Dávila-Félix, 667 F.3d 47, 65 (1st Cir. 2011). But in this case, like in Serrano-Mercado, the government did not introduce any Shepard documents because the defendant did not challenge his classification as an armed career criminal. Thus, there are no Shepard documents in the record on appeal. See Serrano-Mercado, 784 F.3d at 847-48. If we conclude that Massachusetts law is unclear as to whether trafficking is divisible, we will have no way of resolving the divisibility inquiry. Cf. id.

We conclude that the Serrano-Mercado rule extends to these circumstances. On plain error review, a defendant arguing that a prior conviction was improperly counted as an ACCA predicate because there is a non-qualifying form of the offense bears the burden of proving either (1) that the offense is indivisible; or (if the offense is not shown to be indivisible) (2) that the prior conviction was for the non-qualifying form of the offense. Thus, such a defendant can only win on plain error review in the absence of Shepard documents if the prior conviction was for a crime that has a non-qualifying form and its indivisibility can be clearly ascertained without any need to look at She-phard documents as Mathis allowed. 20

We reach this conclusion because abandoning the Serrano-Mercado rule in these circumstances would produce anomalous results. Imagine two defendants, both of whom fail to object to their classification as armed career criminals and both of whom have at least one predicate offense with a non-qualifying form. Neither introduces any Shepard documents at sentencing. Both appeal their sentences. By the vagaries of state law, it turns out that one defendant’s questionable predicate offense was clearly divisible under state law at the time of the conviction, while the other defendant’s questionable predicate offense was not clearly divisible under state law at the time of the conviction. Serrano-Mercado requires that the former bear the burden of proving that his conviction was for the non-qualifying form of the divisible offense using Shepard documents. We think the latter should be likewise required to prove that the offense was not divisible at the time of the prior conviction, also using Shepard documents, under the procedure contemplated in Mathis. We decline to create such an arbitrary distinction.

Having reached 'this conclusion, Bain’s appeal can only succeed on the plain error standard if the Massachusetts trafficking statute was clearly indivisible at the time of his conviction. But we conclude that, based on the arguments made to us in this appeal, the statute’s divisibility is unclear. 21

Bain’s argument that the statute was clearly indivisible rests on several Massachusetts cases, which, he argues, establish that a jury need not be unanimous about the type of trafficking for which a defendant has been convicted. We do not read these cases to establish any clear rule that bears on the statute’s divisibility.

The Massachusetts Supreme Judicial Court has not clearly held whether the different ways of committing the trafficking offense are different crimes with different elements or simply different means of committing a single crime. It has described the statute as “disjunctive, setting forth three categories of trafficking: (1) manufacturing, distributing, or dispensing fourteen grams or more of cocaine; (2) possessing with intent to manufacture, distribute, or dispense fourteen grams or more; and (3) bringing into the Commonwealth fourteen grams or more.” Commonwealth v. Roman, 414 Mass. 642, 609 N.E.2d 1217, 1218 (1993); see also Commonwealth v. Chappee, 397 Mass. 508, 492 N.E.2d 719, 727 (1986). It has not, however, clarified whether this disjunctive form bears on the elements of the offense.

There is some suggestion in Roman that an indictment need only allege “trafficking,” without specifying the form of trafficking. There, the SJC reviewed whether there was sufficient evidence before the grand jury to indict the defendant for trafficking. Id. at 1218. It concluded that there was insufficient evidence for trafficking on an importation theory, but sufficient evidence on a possession with intent to distribute theory. Id. at 1219. The opinion does not quote the indictment, however, so this hardly constitutes a clear holding.

The Massachusetts Appeals Court has also issued several opinions that are ambiguous as to the elements of the offense. Like Roman, one opinion suggests that the statute merely creates one crime—“trafficking”—that can be committed in several different ways. See Commonwealth v. Silva, 21 Mass.App.Ct. 536, 488 N.E.2d 34, 37 (1986). But the court’s only holding in Silva was that the three ways of committing the first type of trafficking-“manufacturing, distributing, or dispensing”—are, in fact, disjunctive. Id. That holding does not help Bain.

Bain also points to several cases holding that evidence adduced at trial or facts admitted in connection with a guilty plea were sufficient to support a conviction for trafficking under multiple theories. See Commonwealth v. Rodriguez, 67 Mass.App.Ct. 636, 855 N.E.2d 1113, 1123 (2006) (sufficient evidence at trial), aff'd 450 Mass. 302, 877 N.E.2d 1274 (2007); Commonwealth v. Panopoulos, No. 99-P-2023, 2001 WL 695106, at *1 (Mass. App. Ct. June 20, 2001) (unpublished disposition) (guilty plea); Commonwealth v. Manrique, 31 Mass.App.Ct. 597, 581 N.E.2d 1036, 1040 (1991) (sufficient evidence at trial). None of these cases demonstrate that the statute was clearly indivisible. In Rodriguez, “[t]he judge instructed the jury on two theories of cocaine trafficking: possession with the intent to distribute, and ‘bringing into’ the Commonwealth.” 855 N.E.2d at 1123 (quoting Mass. Gen. Laws ch. 94C, § 82E(b)(4)). “The judge also told the jurors that they had to be unanimous as to either or both theories, but there was no special verdict form.” Id. The court concluded that there was, nevertheless, no error in submitting a general verdict form to the jury because “the evidence [was] sufficient to sustain a guilty verdict under both theories.” Id. at 1124. Rodriguez points in both directions on divisibility. The trial judge’s instruction that the jury had to be unanimous as to its theory of the case supports the conclusion that the offense was divisible. Nothing in the court’s decision suggests that this instruction was incorrect or that the use of a general verdict form obviated this instruction under state law. The conclusion that there was no error in submitting a general verdict form given that the evidence was sufficient to convict on either theory may cut the other way, but it may also simply be an application of a Massachusetts rule that a general verdict of guilty will stand if there is sufficient evidence for all the theories put before the jury. See, e.g., Commonwealth v. Oquendo, 83 Mass.App.Ct. 190, 982 N.E.2d 538, 542 (2013) (citing Commonwealth v. Plunkett, 422 Mass. 634, 664 N.E.2d 833, 836 (1996)). Rodriguez is simply ambiguous on this point. Manrique and Panopoulos were just sufficiency challenges, not challenges to a verdict form, and are to the same effect. See Manrique, 581 N.E.2d at 1040; Panopoulos, 2001 WL 695106, at *1.

We find these Massachusetts authorities equivocal, in some respects pointing toward indivisibility, in others toward divisibility. The situation strikes us as one in which Mathis suggests the court should consult Shepard documents to determine if the crime is divisible. And the absence of such documents in or as a supplement to the record means that Bain cannot carry his burden on plain error review.

IV.

For the foregoing reasons, we affirm Bain’s conviction and sentence.

1 . The search warrant application included both a search warrant affidavit and the affidavit previously submitted to support the application for the criminal complaint. See United States v. Bain, 155 F.Supp.3d 107, 112 n.1 (D. Mass. 2015). The district court drew from both affidavits in evaluating whether there was probable cause to issue the search warrant. See id. Bain has not objected to this approach.

2 . Connerney identified Bain’s car as a brown Cadillac DeVille with á Massachusetts license plate. The car was registered in Bain's name at an address in Arlington. Connerney confirmed that a police report in a prior case involving Bain reported him driving the same vehicle in July 2011.

3 . We refer to this location as "unit D” rather than "apartment D” because the unit was in fact a condominium, which the tenant rented from someone who owned only that single condominium.

4 . The district court also ruled that no exigent circumstances rendered the entry into unit D reasonable. See id at 122-23. The court therefore did not consider any evidence obtained from that entry when- evaluating the warrant. See id. The government does not challenge this ruling.

5 . We could skip all but the last finding by merely assuming answers to the other issues favorable to Bain. We eschew that more limited approach in order to clarify our case law on the predicate issues in a case in which we have the benefit of very good briefs by both sides. See United States v. Leon, 468 U.S. 897, 924, 104 S.Ct. 3405, 82 L.Ed.2d 677 (1984) ("There'is no need for courts to adopt the ' inflexible practice of always deciding whether the officers’ conduct manifested objective good faith before turning to the question whether the Fourth Amendment has been violated. Defendants seeking suppression of the fruits of allegedly unconstitutional searches' or seizures undoubtedly raise live controversies which Art. Ill empowers federal courts to adjudicate.”). .

6 . We do not rely on the statement in United States v. Cruz Pagán, 537 F.2d 554 (1st Cir. 1976), that the curtilage of each unit in a condominium complex extends only to areas in the exclusive control of the unit’s occupant. Id. at 558. This reasoning was unnecessary to Cruz Pagán's holding, which only concerned whether the resident of a condominium complex has a reasonable expectation of privacy, in the complex’s shared underground garage. ⅛ at 557 (“The legal question which we must resolve is whether the agents' entry into the garage defeated the reasonable expectation of privacy of any of the appellants.”). The court only raised the issue of curtilage- on the ”[a]s-sum[ption] that concepts of curtilage have some relevancy to the Katz inquiry,” id at 558, and, even then, the only conclusion necessary to the court’s holding was that the garage was not part of the unit's curtilage. The broader language about whether other portions pf the condominium complex were within the unit's curtilage was therefore dicta. On top of that, Cruz Pagán dates from before Dunn and did not consider any of the factors deemed significant by Dunn. For these reasons, we do not think Cruz Pagán's broad declaration about tire scope of a condominium unit's curtilage binds us.

7 . Thus, we need not address the government's argument that unit D’s tenant lacked a pos-sessory interest in the door of the unit.

8 . We focus on Concepcion because neither Thompson nor Alvarez adds anything of relevant substance to Concepcion's analysis.

9 . We do not consider whether the curtilage of unit D extended to the entire second-floor landing, which might mean that trying the key on the door to the neighboring apartment was a search of. unit D, or to the entire common space of 131 Laurel Street, which might mean that trying'the key on the door of both of the other apartments in the building were searches of unit D. As we explain in footnote 10, infra, we reject on other grounds the independent soiirce doctrine argument to which these other searches might have been relevant.

10 . At oral argument, the government belatedly sought to argue that by turning the key in the other two units accessible through the front entry, the officers generated information from which the magistrate could infer that the key fit the remaining unit (unit D) even without taking into account the results of turning the key in the lock on unit D. See United States v. Dessesaure, 429 F.3d 359, 367 (1st Cir. 2005), Apart from coming too late, this argument also fails because the affidavit used to secure the warrant contained no information from which the magistrate could determine that only three of the building’s four units were reachable through that entryway. See Whiteley v. Warden, Wyo. State Penitentiary, 401 U.S. 560, 565 n.8, 91 S.Ct. 1031, 28 L.Ed.2d 306 (1971) ("Under the cases of this Court, an otherwise insufficient affidavit cannot be rehabilitated by testimony concerning information possessed by the affi-ant when he sought the warrant but not disclosed to the issuing magistrate.”).

11 . This circuit has left open the question as to whether, in the absence of binding in-circuit precedent, law enforcement may reasonably rely on out-of-circuit case law as providing sufficiently clear precedent. See Sparks, 711 F.3d at 63. This case does not provide an occasion to answer that question because the government argues only that the officers reasonably relied on Lyons and Hawkins.

12 . In fact, only three units were accessible through the front door, but this fact, although helpful to establishing probable cause, was not clearly set forth in the affidavits used to secure the search warrant, Whether we could rely on such an undisclosed but favorable fact given the absence of any motive to conceal it, we need not decide.

13 . Although the affidavits do not make this negative claim, we may infer it from the affidavits' failure to report finding any evidence of this sort on Bain or in his car.

14 . Since the apartment was leased to a person other than Bain, and since officers remained at 131 Laurel Street while Connerney sought the warrant, we cannot presume that anyone must have used the keys to re-enter unit D.

15 . Bain's trial counsel chose not to cross-examine this witness.

16 . He also argues that the fact of a prior conviction must be found by a jury beyond a reasonable doubt. He acknowledges that we are bound by precedent to reject this argument, see Almendarez-Torres v. United States, 523 U.S. 224, 226-27, 118 S.Ct. 1219, 140 L.Ed.2d 350 (1998), and raises it only to preserve it for farther review.

17 . The government árgues that Bain waived—and did not merely forfeit—his argument that he should not have been subject to the ACCA mandatory minimum. We do not decide this question because we assume, favorably to the defendant, that the plain error standard applies.

18 . There is a strong reason to think he is indeed correct under our precedent. See United States v. Mulkern, 854 F.3d 87, 96-97 (1st Cir. 2017).

19 . This rulé is subject to an exception not applicable here. See Serrano-Mercado, 784 F.3d at 849.

20 . This burden-shifting rule retains the exception for the circumstances outlined in Serrano-Mercado. See 784 F.3d at 849.

21 . Neither the defendant nor the government has pointéd us' toward any model jury instructions for this 'statute.