Dissenting opinion filed by Senior Circuit Judge WILLIAMS.
GRIFFITH, Circuit Judge:
Carlton Blount is a prisoner convicted of two murders who filed a habeas petition under 28 U.S.C. § 2254. The district court dismissed the petition as time-barred, a decision that Blount now seeks to appeal. But the federal habeas statute restricts such appeals. Blount’s “appeal may not be taken to the court of appeals” unless he obtains a certificate of appealability (COA). 28 U.S.C. § 2253(c)(1). To do so, Blount must make a sufficient initial showing to a judge or circuit justice. Id. § 2253(c)(l)-(2). Blount requests a COA from us, but because he has not made the necessary showing, we must deny his request.
I
On the night of February 8, 2000, Carlton Blount and two friends attended a basketball game at Blount’s former high school. There, Blount and one of his friends got into a fight with a student at the school, Andre Wallace. After onlookers broke up the fight, Blount and his friends left the school and went looking for Wallace. They eventually drove to the home of Wallace’s girlfriend, Natasha Marsh, and the couple arrived shortly afterwards. An argument erupted, gunshots ensued, and Wallace and Marsh were killed.
Blount was charged in D.C. Superior Court with several counts, including the murders of Wallace and Marsh. At trial, the government’s primary theory was that Blount alone shot the victims. Blount’s defense was that one of his friends pulled the trigger. The judge instructed the jury that even if a friend was the shooter, Blount could be convicted under the doctrine of accomplice liability. The judge’s instructions included the following statements:
It is not necessary that [Blount] have had the same intent that [the] principal offender had when the crime was committed, or that he had intended to commit the particular crime committed by the principal offender.
An aider and abetter is legally responsible for the acts of other persons that are the natural and probable consequences of the crime in which he intentionally participates.
J.A. 121-22. The defense objected that making Blount responsible for the “natural and probable consequences of the crime” allowed him to be convicted even if he lacked the necessary mens rea. Under the judge’s instruction, an accomplice could be convicted of first-degree murder without proof that he acted with premeditation and deliberation. That, the defense argued, eliminated the mens rea element of the offense, violating the Sixth Amendment’s guarantee of a jury verdict on every element. The judge overruled the objection, [735]*735and in February 2001, the jury convicted Blount of first-degree murder of Marsh, second-degree murder of Wallace, and several related counts. Blount was sentenced to prison for sixty-four years to life.
On appeal, Blount was represented by new counsel, who failed to renew the constitutional challenge to the jury instruction. The D.C. Court of Appeals affirmed Blount’s convictions and sentence. It denied his motion for rehearing on October 13, 2004, and formally ended the appeal by issuing a mandate on October 21, 2004. Blount did not petition the U.S. Supreme Court for review.
About two years later, in an unrelated case, the D.C. Court of Appeals held unconstitutional the same jury instruction used in Blount’s case, on the very grounds that Blount had raised at trial. See Wilson-Bey v. United States, 903 A.2d 818, 826, 829-44 (D.C. 2006) (en banc). But because Blount’s conviction had already been affirmed on direct review, the decision had no immediate effect on him.
Meanwhile, Blount had begun a series of unsuccessful pro se collateral challenges to his conviction. On November 16, 2005— over a year after the D.C. Court of Appeals denied his motion for rehearing, the last decision in his direct appeal—Blount signed and mailed from prison to the court a motion for collateral review under D.C. Code § 23-110.1 The motion included a challenge to the accomplice-liability instruction, but the trial court held the claim procedurally defaulted because Blount had failed to raise it on direct appeal. The court denied Blount’s motion and the D.C. Court of Appeals affirmed, issuing its mandate on December 2, 2009.
Blount’s next collateral challenge was a federal habeas petition, which he signed on March 17, 2011.2 Among other arguments, Blount claimed for the first time that he received ineffective assistance of appellate counsel (IAAC) during his direct appeal because his appellate lawyer had failed to challenge the constitutionality of the now-discredited jury instruction. The district court dismissed the petition, holding that Blount had not yet exhausted his local remedies on the IAAC claim, and that the court lacked jurisdiction over his other claims. Blount v. Wilson, No. 11-0743, 2011 WL 1526945 (D.D.C. Apr. 19, 2011). Both the district court and this court denied Blount’s request for a COA. Blount v. Wilson, No. 11-7060 (D.C. Cir. Sept. 27, 2011).
Blount then returned to the D.C. Court of Appeals to raise his IAAC claim through the proper local mechanism: a motion to recall the mandate that the court had issued in his direct appeal. See Watson v. United States, 536 A.2d 1056, 1060 (D.C. 1987) (en banc) (establishing the motion to recall the mandate as the D.C. procedure for raising an IAAC claim). Although Blount’s motion, filed by mail on October 28, 2011, was untimely, the D.C. Court of Appeals excused that flaw and ordered the government to respond on the merits. The court ultimately denied Blount’s motion in late 2012, holding that “based upon the entire record, any error was harmless,” and therefore Blount had “not met the high standard necessary to recall the mandate.” Blount v. United States, No. 01-CF-974, at 1 (D.C. Oct. 11, 2012).
Around the same time as Blount made that motion to recall the mandate, he filed [736]*736a second motion for collateral review under D.C. Code § 23-110, raising claims not directly at issue here. The trial court denied the motion and the D.C. Court of Appeals affirmed in May 2013.
Which brings us to Blount’s present federal habeas petition, filed under 28 U.S.C. § 2254 on September 18, 2013. The rules for habeas petitions challenging state-court judgments govern Blount’s habeas petition, because the federal habeas statute “recognizes that ‘a court of the District [of Columbia] is a state court.’ ” Head v. Wilson, 792 F.3d 102, 106 n.3 (D.C. Cir. 2015) (quoting Madley v. U.S. Parole Comm’n, 278 F.3d 1306, 1308 (D.C. Cir. 2002)). The district court granted the government’s motion to dismiss, holding in relevant part that Blount’s habeas petition was filed outside the one-year limitations period imposed by 28 U.S.C. § 2244(d)(1), enacted as part of the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA), Pub. L. No. 104-132, § 101,110 Stat. 1214,1217. Blount filed a motion for reconsideration under Federal Rule of Civil Procedure 59(e), which the district court denied, holding again that his petition was time-barred.
Blount timely filed a notice of appeal but failed to obtain a COA from the district court. He now requests one from us. We have jurisdiction over that request under 28 U.S.C. § 2253(c), and we appointed amicus curiae to argue in support of Blount.
II
The procedural history of this case is knotted, and the creative theories raised in Blount’s favor only compound that complexity. But the case is complicated, not close. When the knots are untangled and the arguments unpacked, there is no reasonable dispute: Blount’s habeas petition was untimely. The answer is thus clear even under the “limited” inquiry we conduct when considering a request for a COA. Buck v. Davis, — U.S. -, 137 S.Ct. 759, 774, 197 L.Ed.2d 1 (2017).
To obtain a COA, the applicant must make “a substantial showing of the denial of a constitutional right.” 28 U.S.C. § 2253(c)(2). A “substantial showing” is a demonstration “that reasonable jurists could debate whether ... the petition should have been resolved in a different manner or that the issues presented were ‘adequate to deserve encouragement to proceed further.’ ” Slack v. McDaniel, 529 U.S. 473, 484, 120 S.Ct. 1595, 146 L.Ed.2d 542 (2000) (quoting Barefoot v. Estelle, 463 U.S. 880, 893 n.4, 103 S.Ct. 3383, 77 L.Ed.2d 1090 (1983)). Where, as here, “the district court denies a habeas petition on procedural grounds without reaching the prisoner’s underlying constitutional claim,” the successful COA applicant must show us that “jurists of reason would find ... debatable” not only “whether the petition states a valid claim of the denial of a constitutional right,” but also “whether the district court was correct in its procedural ruling.” Id. at 478,120 S.Ct. 1595.
“Where a plain procedural bar is present and the district court is correct to invoke it to dispose of the case, a reasonable jurist could not conclude either that the district court erred in dismissing the petition or that the petitioner should be allowed to proceed further.” Id. at 484, 120 S.Ct. 1595. Because Blount’s petition was plainly time-barred and no jurist of reason could take issue with that procedural ruling by the district court, we deny his request for a COA without having to reach his constitutional claim.3
[737]*737A
Blount’s habeas petition is subject to a one-year limitations period. See 28 U.S.C. § 2244(d)(1). All agree that Blount’s one-year clock runs from the date when his judgment of conviction “became final by the conclusion of direct review or the expiration of time for seeking such review.” Id. § 2244(d)(1)(A). Statutory tolling pauses the clock while “a properly filed application for State post-conviction or other collateral review with respect to the pertinent judgment or claim is pending.” Id. § 2244(d)(2). In some circumstances, equitable tolling can also apply. See Holland v. Florida, 560 U.S. 631, 634, 130 S.Ct. 2549, 177 L.Ed.2d 130 (2010).
Blount’s limitations period began to run on January 11, 2005, when time expired for seeking Supreme Court review in the direct appeal of his conviction. See J.A. 203 (showing that the D.C. Court of Appeals denied rehearing on October 13, 2004); SUP. CT. R. 13 (providing that the time for petitioning for a writ of certiorari expires ninety days after the denial of rehearing). Three hundred nine days of the one-year limitations period then elapsed before Blount filed a local post-conviction motion under D.C. Code § 23-110 on November 16, 2005, starting statutory tolling. That tolling ended on December 2, 2009, when the D.C. Court of Appeals issued its mandate in the section 23-110 proceedings, cf. Holland, 560 U.S. at 638, 130 S.Ct. 2549 (ending statutory tolling when mandate issued), and the limitations period’s remaining 56 days expired on January 27, 2010. Blount filed his present habeas petition in 2013: over three years later, and thus considerably out of time.
B
The recounting of that timeline provides enough analysis to end the case. Blount and amicus raise several alternative theories in an effort to argue that it is at least debatable that his habeas petition was timely. None succeeds.
Principally, Blount and amicus contend that the D.C. Court of Appeals reopened direct review in 2012, thereby restarting the entire one-year limitations period for Blount’s habeas petition that expired in 2010. Under Jimenez v. Quarterman, 555 U.S. 113, 129 S.Ct. 681, 172 L.Ed.2d 475 (2009), a “conviction [is] no longer final for purposes of § 2244(d)(1)(A)” when a court “reopen[s] direct review of [the] conviction.” Id. at 120,129 S.Ct. 681. In that event, the AED-PA limitations clock starts anew when the reopened appeal reaches a final judgment. See id. According to Blount and amicus, the D.C. Court of Appeals reopened direct review in Blount’s case in evaluating his motion to recall the direct-appeal mandate, which Blount made to exhaust his local remedies on his IAAC claim.
That theory fails, however, for the simple reason that the D.C. Court of Appeals never granted Blount’s motion to recall the mandate. There is no reasonable debate over whether the motion was granted: the court “ordered that the motion to recall mandate [be] denied.” Blount v. United States, No. 01-CF-974, at 1 (D.C. Oct. 11, 2012) (capitalization altered). And there is no reasonable debate over whether this disposition reopened the appeal: the D.C. Court of Appeals has never held that denying a motion to recall the mandate reopens an appeal.
[738]*738 To fully explain these issues requires a bit more discussion. When a mov-ant raises an IAAC claim through a motion to recall the mandate, the D.C. Court of Appeals first determines whether the motion has “on [its] face sufficient merit” to justify disturbing the judgment. Watson v. United States, 536 A.2d 1056, 1060 (D.C. 1987) (en banc). If so, the court will then “recall[] the mandate and reopen[] the movant’s appeal in order to fully explore and then decide whether there was ineffective assistance of counsel on the first appeal.” Id. at 1061. By reopening the direct appeal, a recall of the mandate resets the filing deadline for a federal habeas petition, regardless of whether the reopened appeal is ultimately resolved in the appellant’s favor. But critically, the appeal is not reopened until the motion to recall the mandate is granted. Id.; see also id. (“A motions panel granted the request [to recall the mandate], vacated our original judgment, and thereby revitalized [the] direct appeal.” (emphasis added)); Head v. United States, 626 A.2d 1382, 1384 (D.C. 1993) (“[B]y granting the motion to recall the mandate, the direct appeal is revived ....”). This requirement ensures that “judgments will not be disturbed” except “in the presence of exceptional circumstances.” Watson, 536 A.2d at 1060. In Blount’s case, the D.C. Court of Appeals “ordered that the motion to recall mandate [be] denied,” Blount, No. 01-CF-974, at 1 (capitalization altered) (emphasis added), which left the original mandate intact, the direct appeal closed, and the habeas filing deadline untouched.
Blount and amicus try several paths around this obstacle, but each finds a dead end. To begin with, they argue that the D.C. Court of Appeals followed an unconventional procedure in which the court simultaneously denied Blount’s motion to recall the mandate, and reopened and rejected his direct appeal. The D.C. Court of
Appeals followed a similar procedure in Long v. United States, 83 A.3d 369 (D.C. 2013). There, the court granted a motion to recall the mandate raising an IAAC claim and, in the same opinion, resolved the direct appeal by vacating the sentences at issue. Id. at 378, 384. The court explained that it was “combining] those procedural steps”—that is, “resolution of the question of the motion’s merit and resolution of [the] appellant’s re-opened appeal”—because they “require[d] an examination of the same issues, so that a ruling on one [was] essentially a decision on the other.” Id. at 378 n.15. Blount and amicus argue that the D.C. Court of Appeals did the same in his case. In one fell swoop, they argue, the court resolved Blount’s motion to recall the mandate and his underlying IAAC claim. By reaching the latter, they argue, the court reopened direct review of Blount’s appeal.
First, this argument misreads Long. That case, where the court granted the motion, tells us nothing new about this case, where the court denied the motion. Long’s streamlined approach is perfectly consistent with the rule that a motion to recall the mandate reopens an appeal only when granted. In Long, the court first “grant[ed] appellant’s motion to recall the mandate as sufficiently meritorious, and after re-opening appellant’s direct appeal,” then resolved the IAAC claim. Id. at 384 (emphases added). Long’s innovation was to take those two steps in a single opinion. As Blount and amicus would have it, however, Long also made the two steps equivalent: the court’s decision on the motion to recall the mandate becomes the same as a decision on the merits of the IAAC claim in the reopened direct appeal. Not so. Recalling the mandate is antecedent, logically and procedurally, to resolving the direct appeal. Taking both steps in the same opinion made sense in Long, because re[739]*739calling the mandate at the first step allowed it to reach the second step of resolving the appeal, and the court saw no need to wait for a second opinion to do so. When the court refuses to recall the mandate, however, it has no occasion to proceed to the reopened appeal. Put simply, in a single gesture the court can unlock the door and step into the room, but if it leaves the door shut, it can go no further.
Second, even if Blount and amicus were right that D.C. law somehow allows a court to resolve the underlying IAAC claim even when declining to recall the mandate, the D.C. Court of Appeals did not do that here. The court concluded only that Blount “ha[d] not met the high standard necessary to recall the mandate.” Blount, No. 01-CF-974, at 1 (emphasis added). The court did not go beyond that threshold inquiry and take up whether Blount met the standard for vacating his conviction on direct review. According to amicus, there are signs that the court did actually reach the underlying appeal: for instance, the court invoked the merits of Blount’s IAAC claim, based its decision “upon the entire record,” id., and issued a separate order requesting a response from the government. But those signs do not contradict the court’s declaration that it applied only the standard for recalling the mandate; instead, they are fully consistent with that disposition. The movant’s “heavy initial burden” requires him to make “a persuasive ease for recall of the mandate” by showing that the IAAC claim has “sufficient merit” to warrant reopening the appeal and giving the claim full consideration. Watson, 536 A.2d at 1060. To determine whether Blount’s IAAC claim had sufficient merit, the court assessed it in light of the record and the government’s arguments. Concluding that Blount failed to satisfy his initial burden, the court never moved beyond that threshold inquiry.
Having failed to - show that the D.C. Court of Appeals recalled the mandate by using the procedure followed in Long, Blount and amicus take a different tack. They argue that the court’s analysis in denying the motion to recall the mandate was so much like what the court would have done on direct review that the court effectively reopened direct review. In particular, they note that the court denied Blount’s motion on the ground that his IAAC claim lacked merit, which is the same analysis the court would have conducted if reviewing a reopened direct appeal. With no direct support for this notion of functional equivalence, Blount and ami-cus extrapolate from the Supreme Court’s statement in Jimenez that when a court reopens an appeal, the “conviction [is] again capable of modification through direct appeal to the state courts and to [the Supreme Court] on certiorari review.” 555 U.S. at 120, 129 S.Ct. 681. Blount and amicus argue that the D.C. Court of Appeals functionally treated Blount’s conviction as “capable of modification” by assessing the merits of his IAAC claim.
We reject this argument, too. The D.C. Court of Appeals does not reopen a completed appeal simply because its reasoning in denying a motion to recall the mandate mirrors the reasoning it would use on a reopened appeal. That logic leads to incongruous results. Consider a hypothetical motion to recall the mandate raising a wholly frivolous IAAC claim. The court would deny the motion because the IAAC claim is frivolous. If the court were reviewing a reopened direct appeal with the same claim, the court would likewise conclude that the claim is frivolous. According to the argument advanced by Blount and amicus, because the court performed the same analysis when reviewing the motion to recall the mandate as it would have if reviewing a reopened direct appeal, resolving the motion reopened direct review. But [740]*740that would mean that any habeas petitioner could unilaterally reopen direct review and restart his limitations clock simply by filing a frivolous motion to recall the mandate. That cannot be. Instead, the court must deliberately agree to reopen direct review. Here, the court refused to do so.
Jimenez does not suggest otherwise. The passage on which Blount and amicus rely simply describes what happens after a state court issues an “order ... ‘re-stor[ing] the pendency of the direct appeal.’” 555 U.S. at 120, 129 S.Ct. 681 (quoting Ex parte Torres, 943 S.W.2d 469, 472 (Tex. Crim. App. 1997)). After such an order, the conviction becomes “again capable of modification through direct appeal to the state courts and to [the Supreme Court] on certiorari review.” Id. Here, the D.C. Court of Appeals issued no order restoring the direct appeal. The order instead denied the motion to recall the direct-appeal mandate. Blount’s conviction thus never became capable of modification “through direct appeal to” the D.C. courts or the U.S. Supreme Court. Those courts could hear no new direct appeal while the original direct-appeal mandate remained in effect.
In short, the D.C. Court of Appeals did not reopen Blount’s direct appeal because the court never recalled the mandate.4 Our dissenting colleague would resolve the issue differently, but we share a key premise. He concludes that the D.C. Court of Appeals reopened Blount’s appeal because its “disposition of Blount’s motion comprised both a decision to recall and a decision on the merits.” Dissenting Op. at 751. We agree that would be the disposition necessary to reopen the appeal. But we struggle to understand how the court could have “deci[ded] to recall,”s id., when it expressly “ordered that the motion to recall mandate [be] denied,” Blount, No. 01-CF-974, at 1 (capitalization altered). Because we see no reasonable debate over whether the D.C. Court of Appeals denied Blount’s bid to reopen his appeal, he is not entitled to a COA on this theory.
In the alternative, Blount argued to the district court that he is entitled to equitable tolling that would make his habe-as petition timely. Although he does not press that argument before us in his pro se briefs, court-appointed amicus does. Cf. Bowie v. Maddox, 642 F.3d 1122, 1135 n.6 (D.C. Cir. 2011) (“We have no qualms about addressing an argument raised by court-appointed amicus curiae and not by the pro se party on whose behalf he was appointed to present arguments.”). Ami-cus’s first theory of equitable tolling is that Blount’s first federal habeas petition should have been stayed and held in abeyance, an approach endorsed in Rhines v. Weber, 544 U.S. 269, 125 S.Ct. 1528, 161 L.Ed.2d 440 (2005). Under that theory, because the limitations clock would have stopped while the first petition was held in abeyance, equitable tolling would apply from the filing of Blount’s first petition until the filing of the present one. Second, amicus argues that Blount’s pursuit of [741]*741judicial remedies through his first petition, although unsuccessful, warranted equitable tolling during the pendency of that petition. Cf. Irwin v. Dep’t of Veterans Affairs, 498 U.S. 89, 96, 111 S.Ct. 453, 112 L.Ed.2d 435 (1990) (“We have allowed equitable tolling in situations where the claimant has actively pursued his judicial remedies by filing a defective pleading during the statutory period .... ”).
But under either theory, equitable tolling begins only when Blount filed his first federal habeas petition. By then the limitations period had already expired. As explained above in section II.A, the limitations period expired on January 27, 2010, and Blount’s first federal habeas petition was filed over a year later, in March 2011.5 Regardless of the legal merits of amicus’s theories, neither theory would start equitable tolling early enough to make Blount’s petition timely.
Amicus’s only response is that we should modify our calculation of the limitations period in two ways, both of which amicus concedes are actually incorrect. Amicus Br. 10 & n.3, 41-42. First, amicus wants us to use June 6, 2005 as the filing date for Blount’s first motion under D.C. Code § 23-110, which starts statutory tolling. In the district court, the parties and the court assumed that was the proper fifing date, but that was wrong twice over. For one, the date appears to be a typographical error meaning to refer to June 5, 2006, the date on which the D.C. Superi- or Court docketed the motion. See J.A. 190. For another, the docketing date is not the filing date. The motion was filed when Blount mailed the motion on November 16, 2005. See S.A. 69-70; Houston v. Lack, 487 U.S. 266, 276, 108 S.Ct. 2379, 101 L.Ed.2d 245 (1988) (holding that a prisoner files a document at the time he “deliver[s] it to the prison authorities for forwarding to the court clerk”). Next, amicus wants us to adopt a second error from the government’s district-court briefing: using October 4, 2010, when the Supreme Court denied certiorari review of Blount’s first section 23-110 motion, as the end date for statutory tolling. But Supreme Court review does not count towards statutory tolling, so the correct end date is December 2, 2009, when the D.C. Court of Appeals issued its mandate. See Lawrence v. Florida, 549 U.S. 327, 329, 127 S.Ct. 1079, 166 L.Ed.2d 924 (2007) (holding that a petition for certiorari does not extend statutory tolling); Holland, 560 U.S. at 638, 130 S.Ct. 2549 (ending statutory tolling when mandate issued).
Only if we adopted both errors would Blount’s first habeas petition appear timely, making amicus’s equitable-tolling theories viable. Amicus argues that we should do so because the government waived any objection. As amicus observes, the government included both errors in its district-court briefing, mistakenly stated in its appellate briefing that Blount’s first federal habeas petition was timely, and failed to respond to amicus’s waiver argument.
We will use the correct dates, despite the government’s repeated failure to argue for them. Cf. Nattah v. Bush, 605 F.3d 1052, 1058 (D.C. Cir. 2010) (“[W]e may affirm the district court’s decision ‘on the basis of any grounds which support it.’” [742]*742(alterations omitted) (quoting In re Swine Flu Immunization Prods. Liab. Litig., 880 F.2d 1439, 1444 (D.C. Cir. 1989))); cf. also Kamen v. Kemper Fin. Servs., Inc., 500 U.S. 90, 99, 111 S.Ct. 1711, 114 L.Ed.2d 152 (1991) (“When an issue or claim is properly before the court, the court ... retains the independent power to identify and apply the proper construction of governing law.”). It is particularly clear that we have no basis to adopt the first error of using June 6, 2005 as the filing date for Blount’s first D.C. Code § 23-110 motion. That date is not only wrong, but it is undisputedly and indisputably so. As explained above, it is an apparent typographical error without a scintilla of support in the record or the law. To treat the section 23-110 motion as filed on June 6, 2005— and to treat Blount’s first habeas petition as timely and his equitable tolling theories as viable—would be to decide the case on facts and law that all know to be false.
Even if a party’s waiver or forfeiture could ever lead a 'reasonable jurist to accept such an erroneous premise, there is no justification for doing so here. Notwithstanding the government’s mistake, Blount had ample notice of the issue and opportunity to argue the point. In fact, it was amicus—raising the equitable-tolling arguments in support of Blount—who candidly identified the date error. Even more significantly, Blount himself introduced the error. His habeas petition, the opening filing in the present litigation, averred that he filed his D.C. Code § 23-110 motion “[o]n June 6, 2005.” J.A. 6. That may well have been an honest mistake, but all the same, the government and district court used that date only after Blount did first. We see no reason why Blount should get the benefit of an error he introduced. Instead, under all these circumstances, we think it obvious that the correct date should apply.
Finally, AEPDA’s waiver rules do not require otherwise. True, under AEDPA a court may not “bypass, override, or excuse [the government’s] deliberate waiver of a limitations defense.” Wood v. Milyard, 566 U.S. 463, 132 S.Ct. 1826, 1830, 182 L.Ed.2d 733 (2012); cf. also id. at 1833-34 (explaining that a court may, however, excuse the unintentional forfeiture of a timeliness defense). But the government has not waived, forfeited, or conceded its defense that Blount’s current petition is untimely. Far from it: the government raised that defense below, prevailed, and now argues for affirmance on that ground. The government merely faded to make a particular argument, based on the correct date calculation, in support of that defense. Nor did the government waive its limitations defense against Blount’s first habeas petition. To be sure, the government wrongly stated in the present proceedings that the first petition was timely. But that statement does not somehow retroactively waive a defense from the prior proceeding.
In sum, the equitable-tolling theories do not justify a COA. The thread by which those theories hang—that we should calculate Blount’s limitations period using a date all agree is wrong—is too thin to make the issue debatable.6
[743]*743HI
We conclude that jurists of reason would not find debatable the district court’s procedural holding that Blount’s petition was time-barred. We therefore deny Blount’s request for a certificate of appealability.
So ordered.