Affirmed by published opinion. Judge WILKINSON wrote the opinion, in which Senior Judge ELLIS joined. Senior Judge HAMILTON wrote an opinion concurring in part and concurring in the judgment.
OPINION
WILKINSON, Circuit Judge:
Plaintiffs Jerry Kaplan and GO Computer sued Microsoft in 2005 for antitrust injuries that allegedly drove GO out of business in 1994. The four year statutory limitations period that Congress specified has long since run, and Mr. Kaplan and GO seek to combine no fewer than four separate tolling arguments to sidestep it. A necessary link in the chain is plaintiffs’ claim that fraudulent concealment delayed the ticking of the clock. Affirming the district court, we hold that Mr. Kaplan and GO were on inquiry notice of their claims as of 1992, when enough red flags had flown that a reasonably diligent person would have investigated and acted.
I.
Twenty years ago, in 1987, Jerry Kaplan founded GO Corporation to bring to market a handheld computer that could be operated by writing directly on its screen with a stylus, and an operating system, “PenPoint,” for use with that or other touch screen computers. The technology was promising and, initially, GO was on the rise, winning funding and plans for a public endorsement from Intel and catching the attention of other major players in the digital market. As part of its business strategy, GO reached out to software developers — Microsoft among them — to persuade them to write programs for Pen-Point. It likewise approached original equipment manufacturers like Compaq, Fujitsu, and Toshiba to get them to build their own stylus-operated computers based on PenPoint.
According to the complaint, Microsoft saw PenPoint as a threat to its dominance of the operating system market and took anticompetitive steps against GO to extinguish the threat. Specifically, Microsoft pressured Intel to withhold its endorsement and funding; forced other hardware and software developers either not to build products for PenPoint or to pay exorbitant licensing fees if they did so; withheld technical information to ensure that Microsoft’s own software would be incompatible [173]*173with PenPoint; and stole GO’s trade secrets to create a faux competitor to Pen-Point, “PenWindows,” that Microsoft would pull from the market as soon as GO was gone. The pressure was too much; GO closed its doors in January 1994, almost fourteen years ago. The complaint adds that even after GO closed, manufacturers sold hardware based on PenPoint through 2002, and Microsoft still markets products based on trade secrets it stole from GO.
GO’s assets passed to EO Corporation, a former GO spinoff that was trying to develop a hybrid between stylus-based computers and cell phones — another technology that would have its day in time. EO ceased development of PenPoint and functionally shut down in July 1994, when its chief shareholder, AT & T, withdrew funding. A few months later, Microsoft withdrew PenWindows from the market. In 1997, EO was formally dissolved and its assets passed to Lucent Technologies, Inc., which owns PenPoint to this day.
This case might have ended there, with no lawsuit at all, but for the Department of Justice’s May 18, 1998, filing against Microsoft. Twice before the federal government had taken action. The Federal Trade Commission had launched a multi-year investigation of Microsoft in the early 1990s, which it ended in 1993 by declining to pursue an antitrust action. The Department of Justice had compelled Microsoft to sign a consent decree in 1994, prohibiting certain restrictive licensing practices. The Department’s latest suit concluded in the district court with a new consent decree in November 2002, see United States v. Microsoft Corp., 231 F.Supp.2d 144 (D.D.C.2002), and concluded on appeal when the U.S. Court of Appeals for the D.C. Circuit affirmed the entry of the decree in June 2004, Massachusetts v. Microsoft Corp., 373 F.3d 1199 (D.C.Cir.2004).
In 2005, Mr. Kaplan, along with a new corporate entity he had recently formed, co-plaintiff GO Computer, Inc.,1 went to Lucent with a proposal: In exchange for 30% of any judgment or settlement GO and Kaplan might win, Lucent would assign its “legal interest in any federal and state antitrust claims” that Lucent “acquired from AT & T” and that AT & T “had previously acquired from GO.” The deal was struck in April 2005. On June 29, 2005 — a day short of a year after the D.C. Circuit affirmed the consent decree in Massachusetts v. Microsoft— GO filed parallel complaints against Microsoft in federal and state court, the one under federal antitrust law and the other under state antitrust law. Microsoft met GO’s federal complaint with a motion to dismiss or (in the alternative) for summary judgment on statute of limitations grounds.
The merits of GO’s antitrust allegations were never at issue in the district court; argument centered wholly on the limitations period. The statute of limitations for federal antitrust claims bars any action “unless commenced within four years after the cause of action accrued,” plus any tolling. 15 U.S.C. § 15b (2000). “Generally, a cause of action accrues and the statute begins to run when a defendant commits an act that injures a plaintiffs business.” Zenith Radio Corp. v. Hazeltine Research, Inc., 401 U.S. 321, 338, 91 S.Ct. 795, 28 L.Ed.2d 77 (1971). All of the injuries alleged in GO’s complaint occurred before GO closed in January 1994, with the exception of GO’s reference to PenPoint licensing into 2002 and trade secret theft even after 2002. Thus the issues before the district court were GO’s four argu-[174]*174merits for why the statute of limitations did not expire in the eleven and a half years between GO’s closing and filing.
First, GO invoked fraudulent concealment doctrine, claiming that Microsoft had effectively hidden its wrongdoing until 2002. See Supermarket of Marlinton, Inc. v. Meadow Gold Dairies, Inc., 71 F.3d 119, 122 (4th Cir.1995) (starting the clock for concealed injuries from the date they are discovered rather than the date they were committed). Second, GO argued that the federal government’s suit against Microsoft tolled the clock from the date of filing, May 18, 1998. See 15 U.S.C. § 16(i) (2000) (tolling the clock on private antitrust suits “during the pendency” of related suits by the federal government and “for one year thereafter”). Third, GO argued that the federal case did not come to an end with the district court’s entry of the consent decree in 2002, but only with the D.C. Circuit’s affirmance of that decree in 2004. See Russ Togs, Inc. v. Grinnell Corp., 426 F.2d 850, 857 (2d Cir.1970) (holding that the “pendency” of a government antitrust action continues through appeal). Fourth, GO claimed that Microsoft continued to commit antitrust violations involving Pen-Point technology through at least 2002. Zenith Radio, 401 U.S. at 338, 91 S.Ct. 795 (stating that each new injurious act in a continuing antitrust conspiracy starts a new limitations clock as to that act).
In June 2006, the district court, construing Microsoft’s motion as one for summary judgment, rejected GO’s arguments. GO Computer, Inc. v. Microsoft Corp., 437 F.Supp.2d 497 (D.Md.2006). As to fraudulent concealment, the court concluded that enough red flags were up by 1992 to put GO, as a matter of law, on inquiry notice of its claims. The limitations period therefore ran out before United States v. Microsoft was ever filed. Id. at 503. In the alternative, the district court reasoned that GO was certainly on inquiry notice as of July 1994, when the Department of Justice and Microsoft filed their first consent decree. Since July 1994 is just a few months short of four years before May 1998, this alternative view did require engaging GO’s United States v. Microsoft tolling argument. But the court held that United States v. Microsoft ended for tolling purposes in 2002, when the district court entered the consent decree, not in 2004, when the appellate court affirmed it. Id. at 504. The tolling period therefore expired in 2003, two years before GO filed.
Finally, as to the continuing violations argument, the district court pointed out that the language assigning to GO “antitrust claims ... that AT & T had previously acquired from GO” plainly excludes any antitrust claim arising after GO closed in 1994. GO’s continuing violations argument thus rested on injuries to Lucent that GO had no right to allege. GO itself had apparently realized the problem, and, after filing its complaint, had signed two amended assignments with Lucent to fix the original assignment’s language. Neither Microsoft nor the district court were told of the two follow-up assignments, or saw any of the three assignments until a hearing in March 2006, when GO’s counsel produced the third assignment — which was undated — without explaining its provenance. Deeply concerned, the district court sanctioned GO’s counsel in its opinion by “strikfing] all the allegations in the amended complaint underlying the claims asserted by GO for any injury allegedly suffered by Lucent.” 437 F.Supp.2d at 507; see also Fed R. Civ. P. 11(b)(3) (requiring that complaints’ “allegations and other factual contentions have evidentiary support”).
The case did not end there. In a motion for reconsideration, GO’s counsel apologized for the two amended assignments and offered to voluntarily dismiss its feder[175]*175al claims for continuing antitrust injuries to Lucent, promising not to seek reinstatement of those claims or to file a new complaint raising them: “The end result will thus be the same — complete dismissal of the federal case. Plaintiffs seek simply to eliminate the finding that counsel violated Rule 11(b)(3).... ” The district court decided to grant the motion, dismissing the Lucent claims without prejudice (and without Rule 11 sanctions) “in light of plaintiffs’ representation that they do not intend to continue to proceed with their action in this court in any event.” GO then filed a notice of voluntary dismissal, but one not confined to the Lucent claims. On its face, the notice appeared to contemplate dismissal of the entire action. A day after receiving it, however, and before Microsoft’s voice was heard, the court issued an order approving the notice.
Microsoft protested. The district court’s final judgment on statute of limitations grounds, Microsoft explained, would have preclusive effect in California as to GO’s state antitrust claims. GO’s voluntary dismissal of the action was, according to Microsoft, an attempt to wipe the slate clean; GO might then go back to California state court representing its own notice of voluntary dismissal as the last word in the federal case, thereby avoiding the district court’s conclusions on the merits of its statute of limitations arguments. Thus Microsoft moved under Federal Rule of Civil Procedure 59 or 60 that the court alter, amend, or vacate its order granting voluntary dismissal, and enter final judgment in accordance with Federal Rule of Civil Procedure 58. The district court obliged, issuing an order rescinding the grant of voluntary dismissal and stating at a hearing that it had made a “mistake when [it] dismissed the action in its entirety.” On October 30, 2006, the district court signed a final judgment dismissing with prejudice “all claims other than the claims based upon antitrust injuries allegedly suffered by Lucent,” and dismissing without prejudice “claims based upon antitrust injuries allegedly suffered by Lu-cent.”
II.
GO raises three arguments on appeal. First, it challenges our jurisdiction, arguing that the district court’s “Final Judgment,” however styled, was not final for purposes of 28 U.S.C. § 1291 (2000) (“The courts of appeals ... shall have jurisdiction of appeals from all final decisions of the district courts of the United States.... ”).2 Second, GO claims that the district court was not entitled to rescind its initial approval of the notice of voluntary dismissal. Finally, on the statute of limitations issue itself, GO urges three of the arguments the district court rejected: the fraudulent concealment argument, the tolling argument based on United States v. Microsoft, and the variation on that tolling argument based on Massachusetts v. Microsoft. GO does not continue to advance its continuing violations argument. Thus, the only antitrust injuries legitimately before us pre-date GO’s closing in January 1994, and the only limitations arguments concern those injuries.
A.
GO’s § 1291 finality argument stems from the fact that the district court’s “Final Judgment” dismissed some of GO’s claims with prejudice, and some without. [176]*176Relying mainly on Domino Sugar Corp. v. Sugar Workers Local Union 392, 10 F.3d 1064 (4th Cir.1993), GO contends that whenever a district court dismisses any claim without prejudice, such that a plaintiff could conceivably “save his action” and carry on in district court “by merely amending his complaint,” id. at 1066-67, the dismissal is nonfinal. Here, GO argues, it could save its action simply by filing an amended complaint based solely on the claims dismissed without prejudice.
Section 1291’s finality rule has never been so rigid. It is a pragmatic rule, and we have interpreted it chiefly to carry out its “twin purposes” of “avoiding] the enfeebling of judicial administration that comes with undue delay” of ongoing district court proceedings and “preserv[ing] the primacy of the district court as the arbiter of the proceedings before it.” MDK, Inc. v. Mike’s Train House, Inc., 27 F.3d 116, 119 (4th Cir.1994) (internal quotation marks omitted). Indeed, Domino Sugar’s significance has been its insistence that appellate panels “evaluate the particular grounds for dismissal in each case” before either permitting or prohibiting appeals from dismissals without prejudice. 10 F.3d at 1066; see, e.g., Chao v. Rivendell Woods, Inc., 415 F.3d 342, 345 (4th Cir.2005) (“Domino Sugar requires us to examine the appealability of a dismissal without prejudice based on the specific facts of the case in order to guard against piecemeal litigation and repetitive appeals.”).
Dismissals without prejudice naturally leave open the possibility of further litigation in some form. What makes them final or nonfinal is not the speculative possibility of a new lawsuit, but that they “end the litigation on the merits and leave nothing for the court to do but execute the judgment.” MDK, Inc., 27 F.3d at 119 (internal quotation marks omitted); accord Hill v. Potter, 352 F.3d 1142, 1144 (7th Cir.2003) (“The test for finality is not whether the suit is dismissed with prejudice or without prejudice.... The test is whether the district court has finished with the case.”).
When the district court dismissed some of GO’s claims without prejudice, it was utterly finished with GO’s case. The claims in question, of course, are those based on injuries to Lucent that GO never had a right to allege, and that now make an indirect appearance not to support a statute of limitations argument, but to undermine the finality of the district court’s judgment. GO escaped Rule 11 sanctions and won dismissal without prejudice by promising never to raise these claims in federal court again. And even if another district court by some chance did allow GO to file a new complaint for the Lucent claims, that case would be based on distinct facts from this one; in no sense would GO have saved this action by amending this complaint. The district court thus rendered a final judgment, and we have jurisdiction to consider it.
B.
GO’s next claim — that the district court, having issued the order approving GO’s voluntary dismissal of the action, could not rescind it — is also geared to extinguishing the district court’s final judgment and restoring GO’s voluntary dismissal. As GO conceives it, the district court could only alter or rescind an order of voluntary dismissal under one of three rules: Federal Rule of Civil Procedure 41(a)(2) (governing voluntary dismissals), 59(e) (governing motions to alter or amend a judgment), or 60(b) (governing motions for relief from a judgment or order). Each of those rules requires a type and degree of justification the district court, according to GO, did not have.
[177]*177As to Rule 41(a)(2), GO argues that under Davis v. USX Corp., 819 F.2d 1270 (4th Cir.1987), courts should generally grant Rule 41(a)(2) motions unless unduly prejudicial to defendants, and the prospect of a second lawsuit in state court cannot qualify as prejudice. Id. at 1273-75. This argument lays the emphasis in the wrong place. The primary force of 41(a)(2) is to empower district courts to exercise discretion over voluntary dismissals: “[A]n action shall not be dismissed at the plaintiffs instance save upon order of the court and upon such terms and conditions as the court deems proper.” While Davis does indeed direct district courts to “focus primarily” on the interests of defendants, id. at 1273, it never suggests limiting 41(a)(2) to exclude all other considerations&emdash;among them, preventing plaintiffs from litigating, losing, and then wiping the slate clean by voluntarily dismissing their action, see RMD Concessions, L.L.C. v. Westfield Corp., 194 F.R.D. 241, 243 (E.D.Va.2000). Nothing in Rule 41 suggests the district court was constrained in its discretion to rescind the order granting GO’s voluntary dismissal.
As to Rule 59(e), GO claims that Hill v. Braxton, 277 F.3d 701, 708 (4th Cir.2002), permits motions to alter or amend judgments only for changes of law, new evidence not available at trial, to correct a clear error of law, or to prevent a manifest injustice. All true. But here, where the district court meant to permit voluntary dismissal for only one claim and, through accident, almost permitted GO to escape judgment entirely, there was both a clear error of law and a risk of manifest injustice.
As to Rule 60(b), it is impossible to see how a rule designed to grant relief from orders or judgments due to “mistake, inadvertence, surprise, or excusable neglect” could fail to apply here, where the district court stated, “I just think I made a mistake.” Everyone makes mistakes. And district courts can correct them no less than others, unburdened by the tripwire formalisms GO urges. Quite apart from any Rule 59 or 60 motion, it is inconceivable that district courts would not possess the authority to correct an error and spare litigants the serious consequences of mistaken judgment.
C.
We arrive, then, at GO’s three statute of limitations arguments: fraudulent concealment, tolling based on United States v. Microsoft, and further tolling based on Massachusetts v. Microsoft. It bears repeating that the four year antitrust limitations period in 15 U.S.C. § 15b generally starts to run when a cause of action accrues, and a cause of action generally accrues when a defendant commits an act that causes economic harm to a plaintiff. Zenith Radio, 401 U.S. at 338, 91 S.Ct. 795. The only injurious acts GO can legitimately allege pre-date its closing in January 1994&emdash;more than four years before the government filed United States v. Microsoft in May 1998. Thus GO’s only hope is to invoke fraudulent concealment doctrine to start the limitations period later than January 1994; if this argument fails, there is no need to reach the others.3
Fraudulent concealment is an “equitable doctrine ... read into every [178]*178federal statute of limitation.” Holmberg v. Armbrecht, 327 U.S. 392, 397, 66 S.Ct. 582, 90 L.Ed. 743 (1946). It does not stop the clock; it moves the clock, starting it from when the wrong was discovered rather than when it was committed. Bailey v. Glover, 88 U.S. (21 Wall.) 342, 349-50, 22 L.Ed. 636 (1874) (“[W]hen the fraud has been concealed, or is of such a character as to conceal itself, the statute does not begin to run until the fraud is discovered....”); Supermarket of Marlinton, Inc. v. Meadow Gold Dairies, Inc., 71 F.3d 119, 122 (4th Cir.1995) (same). But from its earliest days in this country, there was a condition: The doctrine only applies “when there has been no negligence or laches on the part of a plaintiff in coming to knowledge of the fraud.” Bailey, 88 U.S. at 349. This circuit has thus formalized the doctrine into a three-part test: “[A] claimant must establish that (1) the party pleading the statute [of limitations] fraudulently concealed facts which are the basis of a claim, and that (2) the claimant failed to discover those facts within the statutory period, despite (3) the exercise of due diligence.” Pocahontas Supreme Coal Co. v. Bethlehem Steel, 828 F.2d 211, 218 (4th Cir.1987).
Inquiry notice, which charges a person to investigate when the information at hand would have prompted a reasonable person to do so, touches on the diligence requirement of part three. Brumbaugh v. Princeton Partners, 985 F.2d 157, 162 (4th Cir.1993). The district court held that enough bells went off as of 1992 to put GO — specifically Mr. Kaplan — on inquiry notice of its claims.
By 1991 and 1992, Mr. Kaplan knew with some specificity about the array of obstacles Microsoft was allegedly putting in GO’s way. Twice he met with the FTC as part of its Microsoft investigation. The first time, in 1991, an FTC investigator remarked to him, “This looks like a textbook case of abuse of monopoly power.” GO Computer, 437 F.Supp.2d at 501. The second time, in 1992, Mr. Kaplan gave the FTC a declaration reporting specific conversations in which original equipment manufacturers had told Mr. Kaplan that Microsoft’s licensing arrangements made pre-installing PenPoint either prohibitively expensive or altogether prohibited. The declaration also reported “rumors that Microsoft was warning software developers and hardware manufacturers not to do business with GO”; suggested that Microsoft’s PenWindows was based on GO’s trade secrets; and discussed technical information Microsoft was withholding about its own software. The district court noted further that in a book Mr. Kaplan wrote in 1994, he reported specific occasions, prior to his meeting with the FTC, where hardware manufacturers had told him about Microsoft’s restrictive licensing practices. Id. at 501-02. Finally, after seeing a demonstration of PenWindows in 1991, Mr. Kaplan became so suspicious that he went to a law firm to discuss an intellectual property suit against Microsoft (which the law firm thought was strong, though Mr. Kaplan declined to pursue it due in part to the “cost and distraction of a legal battle”). Id. at 500.
There can be no question that this profusion of information was sufficient, as a matter of law, to spur a reasonably diligent person to investigate an antitrust claim. We must tread cautiously. On the one hand, inquiry notice should not “await the dawn of complete awareness.” Brumbaugh, 985 F.2d at 162. Full knowledge often awaits discovery, and the very notion of “inquiry notice” implies something less than that. On the other hand, too sensitive a trigger will have parties rushing to the courthouse at the first hint of suspicion to avoid having their claims go stale. [179]*179What put Mr. Kaplan so plainly on inquiry notice is the multiplicity and specificity of the information he had. Where a plaintiff knows of a pattern of particular actions that a defendant has taken against him, though the pattern’s precise scope might be unclear and its exact legal ramifications uncertain, the plaintiff is on inquiry notice of his claim. And lest this seem too quick a trigger, it bears emphasis that the date of inquiry notice is not a filing deadline. It is only the date on which a cause of action accrues and the four year period allotted by Congress for a plaintiff to investigate begins.
GO tries to blunt the force of what Mr. Kaplan knew in 1992 with two arguments. First, GO claims that red flags alone are not enough to start the limitations period running, for “if reasonable further inquiry would not have revealed the basis for the antitrust claim, the plaintiffs claim is not time-barred.” Supermarket of Marlinton, 71 F.3d at 128. Had Mr. Kaplan reasonably investigated, GO claims, he would not have been able to prove Microsoft’s monopoly power, nor to penetrate the network of nondisclosure agreements keeping Microsoft’s anticompetitive activities secret.
This argument grafts a novel requirement on the traditional equitable terms of fraudulent concealment: Not only must there be reasonable diligence, but also the potential to reveal the fraud in full, outside of discovery. Brumbaugh states the opposite: “Inquiry notice is triggered by evidence of the possibility of fraud, not by complete exposure of the alleged scam.” 985 F.2d at 162. To be sure, a diligent plaintiff need not engage in ceaseless inquiry when reasonable inquiry does not expose grounds for suit. But nothing in Supermarket of Marlinton excuses a negligent plaintiff from the diligence requirement — not even if a fraud is allegedly well-disguised. Fraud by its nature is something perpetrators take pains to disguise, and plaintiffs’ notion that allegedly concealed fraud excuses the need for any diligence on plaintiffs’ part would permit statutory periods to be tolled indefinitely, even when plaintiffs could reasonably be expected to bring suit.
Second, GO claims the information Mr. Kaplan had was ambiguous. True, he knew of an intellectual property violation, but that does not mean he reasonably knew of an antitrust violation. True, he shared suspicions with the FTC, but he also reasonably relied on the FTC’s public decision not to pursue an antitrust action against Microsoft. And while he suspected Microsoft of stealing his trade secrets, he also relied on Bill Gates’s promise, after the two of them exchanged letters on the issue, unequivocally denying wrongdoing. What was reasonable under these circumstances, GO argues, is properly a factual question that belongs to a jury.
These are manufactured ambiguities. The intellectual property violation Mr. Kaplan investigated does not stand alone, and in context, it was one piece of a larger pattern of Microsoft’s alleged anticompeti-tive conduct. The FTC decision took place in 1993 — a year too late for this discussion. And Bill Gates’s denial of -wrongdoing amounts to little; wrongdoing is not a straightforward matter of fact, and it is not fraud to deny it. Pocahontas Supreme Coal, 828 F.2d at 218-19 (“To permit a claim of fraudulent concealment to rest on no more than an alleged failure to own up to illegal conduct upon this sort of timid inquiry would effectively nullify the statute of limitations in these cases.”). Nor is the denial much ground to stand on when Mr. Kaplan has testified that, as of 1991, he viewed Microsoft as a “bunch of thieves.” GO Computer, 437 F.Supp.2d at 501.
[180]*180III.
The injuries at the core of this case were committed, if at all, no less than eleven and closer to fifteen years before GO filed suit. GO has combined a fraudulent concealment argument, a tolling argument based on the filing of United States v. Microsoft, a variation on that tolling argument based on the appeal in Massachusetts v. Microsoft, and, in the district court, a continuing violations claim to stretch and swell the limitations period to roughly three or four times the one Congress specified. To allow this litigation to proceed would simply eviscerate Congress’ intent. There is a place for finality in the law. Defendants are prejudiced when “[mjemo-ries fade, documents are lost, [and] witnesses become unavailable.” Brumbaugh, 985 F.2d at 162. And defendants are entitled to “the security of knowing when legal action against [them] has been foreclosed.” Id. There was a time when these claims could and should have been fairly adjudicated, but that time has long passed. The district court held that “rather than being ignorant of grounds upon which it could have sued Microsoft, GO made business and strategic decisions not to pursue legal action.” 437 F.Supp.2d at 500. This conclusion presents no issue of triable fact, and the judgment is
AFFIRMED.