*700Opinion for the Court filed by Circuit Judge WILLIAMS.
Concurring opinion filed by Circuit Judge BUCKLEY, in which Circuit Judge D.H. GINSBURG joins.
Dissenting opinion filed by Circuit Judge EDWARDS, in which Circuit Judges MIKVA and RUTH BADER GINSBURG and Senior Circuit Judge ROBINSON join.
STEPHEN F. WILLIAMS, Circuit Judge:In Gay Rights Coalition v. Georgetown University, 536 A.2d 1, 39 (D.C.1987), the District of Columbia Court of Appeals held that the University’s refusal to grant the Coalition “equal access to its ‘facilities and services’ ” violated the District’s Human Rights Act. Congress responded with § 145 of the D.C. Appropriations Act of 1989, known in this litigation and in the media as the “Armstrong Amendment” after its prime mover in the Senate. Pub.L. No. 100-462, § 145,102 Stat. 2269, 2269-14 (1988). Section 145(b) provided: “None of the funds appropriated by this Act shall be obligated or expended after December 31, 1988, if on that date the District of Columbia has not adopted subsection (c) of this section.” Subsection (c) — the language specified for addition to the D.C.Code — in turn read: “[I]t shall not be an unlawful discriminatory practice in the District of Columbia for any educational institution that is affiliated with a religious organization or closely associated with the tenets of a religious organization to deny, restrict, abridge, or condition ... the use of any fund, service, facility, or benefit; or the granting of any endorsement, approval, or recognition, to any person or persons that are organized for, or engaged in, promoting, encouraging, or condoning any homosexual act, lifestyle, orientation, or belief.”
Instead of enacting the language, the members of the D.C. Council sued for in-junctive and declaratory relief against enforcement of the funding condition. The district court found that the condition violated their First Amendment rights by coercing them to speak against their wills, and issued a declaratory judgment. Clarke v. United States, 705 F.Supp. 605 (D.D.C.1988). Having given plaintiffs relief on that claim, the court did not reach their other theories. The government appealed, and a panel of this circuit upheld the district court on September 26, 1989. Clarke v. United States, 886 F.2d 404 (D.C.Cir.1989).
Although the 1989 Appropriations Act was due to expire September 30, 1989, four days after the panel decision, the Act was extended several times as the 1990 Appropriations Act wound its way through Congress. The last extension expired on November 20, 1989, Pub.L. No. 101-154, 103 Stat. 934 (1989), and the 1990 Act became law when the President signed it the next day. Congressional Index, 101st Cong. (1989-90) (CCH) at 35,050 (H.R. 3746). The new act did not contain a funding condition; instead, Congress used its power under the District of Columbia Self-Government and Governmental Reorganization Act, Pub.L. No. 93-198, § 601, 87 Stat. 774 (1973), to amend the D.C.Code directly, bypassing the D.C. Council. District of Columbia Appropriations Act, 1990, Pub.L. No. 101-168, § 141, 103 Stat. 1267, 1284 (1989).
In the meantime, the United States had filed a petition for rehearing and suggestion for rehearing en banc. Once the last extension expired and the 1990 Act became law, it filed a motion suggesting that the case was moot and that the panel decision be vacated. On December 15, 1989, both the panel and the full court declined to rehear the case on the merits, but withheld the mandate. Clarke v. United States, 898 F.2d 161 (D.C.Cir.1989). The panel denied the proposed vacatur for mootness, but the full court ordered the mootness issue heard en banc.
We now decide that the issue was moot after the expiration of the last extension of the 1989 Act, and that the proper remedy is vacatur of the panel opinion and a remand to the district court with instructions to vacate its opinion and to dismiss the relevant count of the complaint as moot.
I. Mootness
A. Without regard to the doctrine’s exceptions.
The mootness doctrine, deriving from Article III, limits federal courts to *701deciding “actual, ongoing controversies.’ Honig v. Doe, 484 U.S. 305, 317, 108 S.Ct. 592, 600-01, 98 L.Ed.2d 686 (1988). Even where litigation poses a live controversy when filed, the doctrine requires a federal court to refrain from deciding it if “events have so transpired that the decision will neither presently affect the parties’ rights nor have a more-than-speculative chance of affecting them in the future.” Transwesern Pipeline Co. v. FERC, 897 F.2d 570, 575 (D.C.Cir.1990); see also Preiser v. Newkirk, 422 U.S. 395, 401, 95 S.Ct. 2330, 2334, 45 L.Ed.2d 272 (1975). This limitation “subsists through all stages of federal judicial proceedings, trial and appellate.” Lewis v. Continental Bank Corp., — U.S. -, 110 S.Ct. 1249, 1253, 108 L.Ed.2d 400 (1990).
Appellees do not claim that the Armstrong Amendment has had any residual effect on their First Amendment rights since it was superseded by the 1990 Appropriations Act. The 1989 Act appropriated funds “for the District of Columbia for the fiscal year ending September 30, 1989,” Pub.L. No. 100-462, 102 Stat. at 2269, and § 108 of the 1989 Act pinned down its automatic expiration with an explicit statement that “[n]o part of any appropriation contained in this Act shall remain available for obligation beyond the current fiscal year unless expressly so provided herein.”1 Pub.L. No. 100-462, § 108, 102 Stat. at 2269-8. After the 1989 Act’s last extension expired, the Armstrong Amendment’s funding condition ceased to have any effect.
Appellees argue, however, that the case is not moot because vacating the district court's declaratory judgment would leave open the “formal possibility of prosecution” under the Anti-Deficiency Act, which forbids “knowingly and willfully” expending government funds without authorization from Congress. Brief for Appellees at 39-40; see 31 U.S.C. §§ 1341(a), 1342 & 1350 (1988). They rely primarily on Edgar v. MITE Corp., 457 U.S. 624, 102 S.Ct. 2629, 73 L.Ed.2d 269 (1982), in which the Court found that the possibility of “civil and criminal liability” saved the case from mootness. Id. at 630, 102 S.Ct. at 2634. The plaintiff corporation had made a tender offer in disregard of Illinois’s anti-takeover law and secured an injunction against the law’s enforcement, but it later withdrew the offer. The Secretary of State of Illinois, who was charged with enforcing the Illinois act, had represented to the court of appeals that he “intend[ed] to enforce the Act against MITE.” Id.
Here, by contrast, there is no such threat. Quite the opposite: the government at oral argument not only stated that “no one has ever suggested that there would be [a prosecution],” but also conceded “formally for the record that the existence of a judgment during that time would be a complete and adequate defense to any prosecution.”
The concession gives formality to the obvious — the nonviability of any such prosecution. Our research has failed to turn up a single prosecution under the Anti-Deficiency Act in its entire existence since 1905. 33 Stat. 1214, 1257-58, ch. 1484, § 4 (1905). Moreover, the appellees would have at least two strong legal arguments should some future prosecutor try to dust it off. First, the district court’s decision (and the panel affirmance) would raise a serious question whether appellees had the state of mind necessary for a violation. Second, although the MITE majority declined to resolve whether a later-vacated federal injunction would protect MITE from a state prosecution, 457 U.S. at 630, 102 S.Ct. at 2634, and Justice Stevens concluded that it could not, id. at 647-54, 102 S.Ct. at 2643-47, no federalism concerns would prevent an immunizing effect in the present case.2 *702Indeed, the few circuits faced with the question have held that a federal judgment, later reversed or found erroneous, is a defense to a federal prosecution for acts committed while the judgment was in effect. Thus in United States v. Mancuso, 139 F.2d 90 (3rd Cir.1943), the court held that a draftee who did not report for duty because a district court had enjoined enforcement of his induction order could not be prosecuted for failing to appear — despite the injunction’s having been issued erroneously. And in United States v. Albertini, 830 F.2d 985 (9th Cir.1987), the court reversed a criminal conviction for acts taken in conformity with a decision of its own that was later reversed by the Supreme Court. See also United States v. Brady, 710 F.Supp. 290 (D.Colo.1989); compare United States v. Bruscantini, 761 F.2d 640, 641-42 (11th Cir.1985) (rejecting argument that reliance on stated opinion of state judge and state prosecutor was a defense to federal charge).
Of course we cannot say that the risk of an attempted prosecution is zero. Later government representatives might try to persuade a court to distinguish Mancuso or to reject it. They might also believe that the government’s present representation created no estoppel, and conceivably a court would agree. But see IB Moore’s Federal Practice ¶ 0.405[8] at 239-40 (1988); Farmland Indus., Inc. v. Grain Bd. of Iraq, 904 F.2d 732, 739 (D.C.Cir.1990). But zero risk is not the test. Any assessment of the possible side effects of a case “really should be painted in shades of gray, since few controversies are wholly beyond the power of changed circumstances to revive.” Commodity Futures Trading Comm’n v. Bd. of Trade, 701 F.2d 653, 655 (7th Cir.1983); see also Wright, Miller & Cooper, Federal Practice and Procedure § 3533.3 at 273-74 (1984) (framing issue as whether “the prospect of future effects is too remote to justify decision”); compare Babbitt v. United Farm Workers Nat’l Union, 442 U.S. 289, 302, 99 S.Ct. 2301, 2310-11, 60 L.Ed.2d 895 (1979) (pre-enforcement review of allegedly unconstitutional statute is allowable when “fear of criminal prosecution ... is not imaginary or wholly speculative”). The government’s formal representation, and its stipulation as to the exonerating effect of the district court judgment, seem to us enough to overcome the “formal” — and realistically quite remote — possibility of prosecution.
While an explicit assertion of intent to prosecute can preserve a conflict’s vitality, see MITE, and the risk may be great enough where the record on intent to prosecute is a blank, Charles v. Daley, 749 F.2d 452, 457 (7th Cir.1984) (amendment of criminal statute does not render challenge moot where defendants “offered no evidence to rebut [the] possibility of prosecution” for violations committed while deleted language was in effect),3 we have found no case where a court has found that the possibility of prosecution kept a controversy alive in the face of the government’s formal legal concession of the existence of a complete defense. The closest is Mesa Petroleum Co. v. Cities Service Co., 715 F.2d 1425 (10th Cir.1983), where the court said that a prosecutor’s “absence of ... intent” to prosecute provided “too slender a reed on which to rest our decision as to whether a live controversy is presented.” Id. at 1431. In fact the officer charged with the enforcement of the act had simply told the court he had no current plans to prosecute. Before the district court, the agency had said that “state law is determinative of [its] enforcement intentions,” id., language that may well have seemed a *703deliberate equivocation. These quibblings are a far cry from the government’s representation and concession here.
All told, then, appellees’ proclaimed fear of prosecution under the Anti-Deficiency Act is far too speculative to keep the case alive.
B. Exceptions to Mootness.
Appellees argue that even if the case is superficially moot, it falls within two exceptions to the doctrine — the common one for “actions capable of repetition yet evading review,” and the related but narrower one addressing a defendant’s voluntary cessation of the offending conduct. Both exceptions involve instances where, despite the apparent demise of the controversy, its resolution has a reasonable chance of affecting the parties’ future relations.4
Common to both exceptions is the task of defining the wrong that the defendant is alleged to have inflicted. What is the injury that is capable of repetition, and what are the “old ways” to which the voluntarily ceasing defendant might return? See United States v. W.T. Grant Co., 345 U.S. 629, 632, 73 S.Ct. 894, 897, 97 L.Ed. 1303 (1953) (voluntary cessation exception rooted in interest in protecting plaintiff from defendant’s possible “return to his old ways”). The opportunities for manipulation are great. The more broadly we define the wrongful conduct, the more numerous are the possible examples, and the greater the likelihood of repetition. Of course there are practical limits. As the defined injury is broadened, it is likely to sweep up actions that last a long time and thus do not evade review. Here, for example, if we defined the alleged wrong as conditioning funding upon legislators’ performance of the expressive conduct inherent in legislation, the action would by no means evade review, as Congress often embeds such conditions in permanent legislation. See, e.g., 23 U.S.C. § 158 (1988) (withholding a percentage of federal highway funds from states that do not prohibit those under 21 from buying alcohol); 47 U.S.C. § 399 (1982) (prohibiting any noncommercial educational broadcasting station that receives federal funds from “editorializing” or endorsing candidates).
The United States would define the alleged injury as congressional use of funding to induce appellees to enact specific language allowing religiously affiliated institutions to discriminate against homosexuals. But facts completely irrelevant to any intelligible formulation of plaintiffs’ claim — such as the specific issue on which Congress sought to induce plaintiffs’ votes — are equally irrelevant to the mootness issue. See Amalgamated Transit Union, AFL-CIO v. Brock, 809 F.2d 909, 914-15 (D.C.Cir.1987).
In their brief opposing mootness, appel-lees appear to wobble in their notion of the injury. At one point they frame the issue as “whether Congress can coerce the Council Members’ votes by withholding appropriated funds,” Brief for Appellees at 31, while elsewhere they speak of Congress’s “attempt[ing] to coerce the Council Members into voting to adopt a prescribed law,” id. at 24 (emphasis added), thus seeming to echo their complaint’s focus on Congress’s insistence on enactment of specific language. In any event, where plaintiffs are resisting a mootness claim we think they must be estopped to assert a broader notion of their injury than the one on which they originally sought relief. Cf. Tallahassee Memorial Regional Med. Ctr. v. Bowen, 815 F.2d 1435, 1449-50 & n. 28 (11th Cir.1987) (looking to complaint to determine scope of plaintiff’s alleged injury).
Here the plaintiffs never sought a broad invalidation of conditioned funding generally. Indeed, they made no claim at all against the other provisions in the 1989 Act that conditioned funds on District legislation but refrained from demanding specific language. See, e.g., Pub.L. No. 100-462, *704§ 141, 102 Stat. at 2269-13 (conditioning expenditure of funds on the D.C. government's implementing “a preference system that does not preclude the hiring of noncity residents”); id., § 143, 102 Stat. at 2269-13 (conditioning all funds on the District’s repealing an existing law prohibiting testing for AIDS as.a condition of acquiring insurance, and passing another to allow for such testing); id., Federal Payment to the District of Columbia, 102 Stat. at 2269 (conditioning $430 million on the District’s maintaining a police force of at least 3880 officers).
Instead, appellees attacked the Armstrong Amendment as violating their First Amendment rights by conditioning funds on their enactment of particular language. Count III of the Complaint, the one relevant, here, is titled “Violation of the Free Speech Clause of the First Amendment by Coercing Particular Speech by the Members of the Council.” Joint Appendix 32 (emphasis added). It goes on to argue that “[t]he Armstrong Amendment coerces plaintiffs to propose and to vote in favor of amending the D.C. Human Rights Act, using the particular language dictated by Congress.” Id. at If 61 (emphasis added); see also id. at ¶ 62 (“A law compelling particular speech, especially particular political speech, is presumptively invalid....”).
It may well be appropriate to narrow the injury or conduct still further, by reference to (1) the especially coercive character of Congress’s having conditioned all District appropriations on enactment of the required legislation, and (2) the gratuitous character of its use of conditioned funding to do something it could easily have done by direct exercise of its plenary power over the District. Count III speaks of the Armstrong Amendment’s having “coerced” the plaintiffs into adopting the specified legislation, id., and earlier argues that by “threatening to freeze all District expenditures, the Armstrong Amendment coerces” plaintiffs’ votes, id. at 1133; so perhaps Count III used “coercion]” in that very specific sense. Also, Count III explicitly says that Congress failed to “proceed in a manner designed to constitute the least possible interference with free speech,” id. at ¶ 62, seemingly invoking Congress's alternative option. We need not resolve whether these were essential to plaintiffs’ original theory of their claim, however, as we find below that even the injurious act as more broadly conceived — conditioning funding on enactment of legislation in specific language — is not “capable of repetition yet evading review” and cannot properly be characterized as conduct that the defendants have voluntarily ceased.
Capable of repetition, yet evading review. In order to fit the case into one of the “exceptional situations” to which this doctrine applies, see Continental Bank, 110 S.Ct. at 1255 (quoting Los Angeles v. Lyons, 461 U.S. 95, 109, 103 S.Ct. 1660, 1669, 75 L.Ed.2d 675 (1983)), appellees must demonstrate that “(1) the challenged action is in its duration too short to be fully litigated prior to its cessation or expiration, and (2) there [is] a reasonable expectation that the same complaining party would be subjected to the same action again.” Murphy v. Hunt, 455 U.S. 478, 482, 102 S.Ct. 1181, 1183, 71 L.Ed.2d 353 (1982) (quoting Weinstein v. Bradford, 423 U.S. 147, 149, 96 S.Ct. 347, 348-49, 46 L.Ed.2d 350 (1975)); see also Honig v. Doe, 484 U.S. 305, 318-20, 108 S.Ct. 592, 601-02, 98 L.Ed.2d 686 (1988). Whatever the precise meaning of “reasonable expectation,” compare Honig, 484 U.S. at 318-19 n. 6, 108 S.Ct. at 601-02 n. 6 (1988) (opinion of Brennan, J., for the Court), with id. at 333-36, 108 S.Ct. at 609-11 (Scalia, J., dissenting), appellees have not shown it.
In estimating the likelihood of an event’s occurring in the future, a natural starting point is how often it has occurred in the past. Here, the Armstrong Amendment represents the sole occasion on which Congress has used conditioned funding to induce the D.C. Council (or any other government, so far as appears) to enact particular language into law. The amendment thus was, as a member of this court put it at oral argument, a purple cow. We have no reason at all to expect to see one ever again, especially as Congress is free to *705avoid problems of the present sort by-amending the D.C.Code directly.
It is true that Senator Armstrong stated at one point that he was “tempted to simply offer the same amendment again.” 135 Cong.Rec. S11104 (daily ed. Sept. 14, 1989). The resisted temptations of one senator, however, even if expressed on the floor of Congress, are hardly a good barometer of the likelihood of a congressional repeat.
Further, even if Congress were again to try to induce enactment of precisely worded legislation by means of conditioned funding, the effort would not necessarily evade review, or even be especially likely to do so. Congress often embeds a conditioned funding mechanism in permanent law. See p. 703 above. There is no reason to think that conditions exacting particular language are systematically linked with short-lived appropriations; with only one instance, there is no way to generalize.
Voluntary cessation. Early cases developing the exception for a defendant’s voluntary cessation focused on preventing a private defendant from manipulating the judicial process by voluntarily ceasing the complained of activity, and then seeking a dismissal of the case, thus securing freedom to “return to his old ways.” See, e.g., W.T. Grant, 345 U.S. at 632, 73 S.Ct. at 897; United States v. Trans-Missouri Freight Ass’n, 166 U.S. 290, 308-09, 17 S.Ct. 540, 546-47, 41 L.Ed. 1007 (1897); cf. Goshen Mfg. Co. v. Hubert A. Myers Mfg. Co., 242 U.S. 202, 207-08, 37 S.Ct. 105, 106-07, 61 L.Ed. 248 (1916). See also Wright, Miller & Cooper, Federal Practice and Procedure, § 3533.5 at 326 (1984). Even assuming Congress could properly be classified as a defendant in this action (which we doubt), the argument fails.
Although the doctrine has more recently been applied to legislative bodies, see, e.g., City of Mesquite v. Aladdin’s Castle, Inc., 455 U.S. 283, 289, 102 S.Ct. 1070, 1074-75, 71 L.Ed.2d 152 (1982); Quern v. Mandley, 436 U.S. 725, 733-34 n. 7, 98 S.Ct. 2068, 2074 n. 7, 56 L.Ed.2d 658 (1978), appellees point to no case applying the doctrine to Congress, and we have found none. We have serious doubts whether it could be. At least in the absence of overwhelming evidence (and perhaps not then), it would seem inappropriate for the courts either to impute such manipulative conduct to a coordinate branch of government, or to apply against that branch a doctrine that appears to rest on the likelihood of a manipulative purpose.
Although the Supreme Court seems not to have addressed the issue, we think it telling that it failed even to mention voluntary cessation in a case where, if it encompassed Congress, the doctrine would be highly relevant. In Department of Treasury v. Galioto, 477 U.S. 556, 106 S.Ct. 2683, 91 L.Ed.2d 459 (1986), the Court dismissed as moot a challenge to statutes effectively denying anyone who had been committed to a mental institution access to firearms. While the case was pending before the Supreme Court, Congress amended the law to provide an administrative remedy for those affected. There, as in Quern and City of Mesquite, Congress was free to “return to its old ways” when the action was dismissed.
We need not decide the issue, however, for we find that non-reenactment of a onetime condition that expired of its own terms cannot be viewed as cessation of conduct. In every case applying the voluntary cessation doctrine, the decision to stop the disputed activity was made while the litigation was pending, by, for example, altering an ordinance, City of Mesquite, withdrawing from a federal program, Quern, or resigning from boards of competing companies, W.T. Grant. Here, the expiration date of the 1989 Act was set well before this dispute arose, and was chosen for reasons having nothing to do with the Armstrong Amendment. (Nor, of course, did the brief extensions of the 1989 Act arise from the amendment or its litigation.) In essence, Congress shot an arrow into the air, and it fell to earth. It stretches the words beyond recognition to say that Congress “voluntarily ceased” anything merely because it refrained from shooting some more arrows after the first landed. More important, extension of the *706doctrine to this case would not fit its basic theory. While the cessation of an ongoing activity pending a lawsuit may well imply an intent to renew the activity once the court has dropped out, this is hardly true of Congress’s allowing a one-time provision to pass into history by its own terms.
Appellees make much of a claim that Congress’s non-renewal of the Armstrong Amendment was moved by a purpose to evade judicial review. We will assume ar-guendo that such a purpose played a significant role in the decision. But as Congress possesses an indisputably valid procedure for achieving its substantive purposes (namely, direct amendment of the D.C. Code), its nonrenewal of the disputed procedure hardly suggests either the manipulative purpose, or the risk of recurrence, that drives the voluntary cessation exception. Cf. Flynt v. Weinberger, 762 F.2d 134, 135 n. 1 (D.C.Cir.1985) (declining to apply voluntary cessation doctrine where “the challenged act is short-term in nature and discontinued after having achieved its objective”).
II. Remedy
Having determined that the case is moot, we must still decide whether it is appropriate to vacate the panel opinion and remand to the district court with instructions to dismiss Count III of the complaint. We find that it is.
In United States v. Munsingwear, 340 U.S. 36, 71 S.Ct. 104, 95 L.Ed. 36 (1950), the Court noted that the “established practice of the Court in dealing with a civil case from a court in the federal system which has become moot while on its way here or pending our decision on the merits is to reverse or vacate the judgment below and remand with a direction to dismiss.” Id. at 39 & n. 2, 71 S.Ct. at 106-07 & n. 2 (collecting cases); see also City of Mesquite, 455 U.S. at 288 n. 9, 102 S.Ct. at 1074 n. 9. Federal courts of appeals have followed suit, vacating district court decisions that became moot before the court of appeals’ own disposition. See, e.g., Matson Navigation Co. v. United States, 825 F.2d 502 (D.C.Cir.1987).
The Supreme Court has also vacated its own decision when a case became moot after the decision issued but before its disposition of a petition for rehearing. Stewart v. Southern Ry. Co., 315 U.S. 784, 62 S.Ct. 801, 86 L.Ed. 1190, vacating as moot 315 U.S. 283, 62 S.Ct. 616, 86 L.Ed. 849 (1942). Again the courts of appeals have pursued a parallel course. Thus in United States v. Caraway, 483 F.2d 215 (5th Cir.1973), the court en banc vacated a panel decision because the case had become moot after the panel opinion issued but “prior to the issuance of the mandate of this court and prior to the sua sponte determination to consider the appeal en banc.” Id. at 216. Similarly, in In re Ghandtchi, 705 F.2d 1315, 1316 (11th Cir.1983), a panel vacated its decision when the case became moot after the opinion issued, but before the time limit for seeking en banc ran or the mandate issued, and in United States v. Miller, 685 F.2d 123 (5th Cir. Unit B 1982), a panel vacated its opinion on learning “[bjefore issuance of the mandate” that the case had become moot. In addition, panels vacate their decisions on learning that the case became moot before issuance. Ruiz v. Estelle, 688 F.2d 266 (5th Cir.1982) (partial vacatur, limited to issue parties had settled); Bumpus v. Clark, 702 F.2d 826 (9th Cir.1983). Wright, Miller and Cooper go further. They argue that because the Supreme Court conventionally grants cer-tiorari on moot cases, and vacates and remands with instructions to dismiss, it is “appropriate for a court of appeals to vacate its own judgment if it is made aware of events that moot the case during the time available to seek certiorari.” Federal Practice and Procedure § 3533.10 at 435.
As this case became moot on November 20, 1989 with the lapse of the 1989 Appropriations Act, and there was on that date no disposition of the suggestion for en banc review and no issuance of the mandate, the standard practice of both the Supreme Court and the courts of appeals calls for automatic vacatur.
The present case presents, however, a curiosity, in that the court’s decision of December 15, 1989 not to rehear the merits *707en banc could be viewed as removing the prime reason for vacatur — to protect the losing party from the collateral effects of a judgment that it might have been able to have overturned but for the mooting event.5 Munsingwear, 340 U.S. at 40-41, 71 S.Ct. at 107.
But this sequence does not affect the matter. As the court can dis-ew banc a case that it has ordered heard en banc, see Bartlett v. Bowen, 824 F.2d 1240 (D.C.Cir.1987), it presumably retains the power to reverse a denial of en banc review until the issuance of the mandate. Indeed, in rejecting the call for rehearing en banc, the court explicitly withheld issuance of the mandate “pending disposition by the en banc Court of appellant’s Suggestion of Mootness and Motion to Vacate.” Clarke v. United States, 898 F.2d 161 (D.C.Cir.1989). At a minimum, our withholding of the mandate must insulate the mootness and vacatur issues from the merits vote; the outcome on the merits cannot moot the mootness.
Appellees suggest two reasons not to vacate: first, that vacatur of a decision that becomes moot after issuance is inappropriate where mootness results from a deliberate act of the losing party rather than “by happenstance”; second, that vaca-tur of a decision moot after issuance is discretionary, not mandatory, and that our discretion should be exercised against vaca-tur. We can resolve these claims without trying to speak definitively to the scope of the first or the soundness of the second. The first is inapplicable; and to the extent the issue is discretionary, we believe the context argues for vacatur.
In United States v. Garde, 848 F.2d 1307 (D.C.Cir.1988), we declined to vacate under Munsingwear where a government agency appealed a district court decision, and then complied with its requirements and applied for its vacatur. We reasoned that a practice of vacatur on these facts would encourage litigants who were dissatisfied with a result to wipe out its precedential and preclusion effects by compliance, and would be unfair to winners in the district court by cheating them of some of the fruits of their victory. See also Center for Science in the Public Interest v. Regan, 727 F.2d 1161 (D.C.Cir.1984); but see Board of Regents v. New Left Project, 414 U.S. 807, 94 S.Ct. 118, 38 L.Ed.2d 43 (1973) (vacating without explanation, 472 F.2d 218 (5th Cir.1973), and remanding to district court for vacatur of its judgment; the court of appeals had refused to vacate district court judgment on Garde-\ik& grounds).
Some commentators have argued that courts can better address these concerns in determining whether the voluntary cessation of conduct by the defendant has caused the case to become moot in the first place. See Wright, Miller & Cooper, Federal Practice and Procedure § 3533.10 at 430-31 (1984). Whatever the merits of that view, it exposes the point that these cases are only applying a milder version of the voluntary cessation doctrine: the case is not live enough for adjudication on the merits, but because of fears of manipulation by the losing party the court denies it the benefit of vacatur. See Commodity Futures Trading Comm’n v. Board of Trade, 701 F.2d 653, 657 (7th Cir.1983) (noting the link). For the reasons developed at pp. 705-06, however, no version of the voluntary cessation doctrine is applicable here. Even if Congress were the defendant, its non-reenactment of a one-time provision that expired pursuant to its own terms — terms adopted in advance of any litigation and with no anticipation of its arising — cannot be considered voluntary cessation or “compliance” with the panel decision.
Two recent decisions have suggested that where a court of appeals decision is mooted after issuance, the decision to vacate should be discretionary. Armster v. United States Dist. Court, 806 F.2d 1347, *7081355 (9th Cir.1986) (although finding the case not moot, id. at 1353-54, and subject to one of the exceptions to mootness, id. at 1357-61, court argues that where mootness follows decision vacatur is within court’s discretion); Commodity Futures Trading Comm’n (focusing exclusively on issue preclusion effects, court declines to vacate district court decision or appellate decision issued before mooting event).
Whatever the merits of this view — which would appear to represent a change from Stewart v. Southern Ry. Co., 315 U.S. 784, 62 S.Ct. 801, 86 L.Ed. 1190 (1942) — the judges of the majority who believe it correct also believe that on the present facts vacatur is appropriate for both our and the district court decisions.
Any decision granting the plaintiffs relief on their First Amendment theory, no matter how narrowly confined (e.g., limited to cases conditioning 100% of funds available to the governing entity — including those from sources other than the federal government — and requiring enactment of specific language), would carry broad implications. Except for Miller v. Town of Hull, 878 F.2d 523 (1st Cir.1989), the panel opinion is the sole court of appeals decision finding legislators’ votes subject to conventional First Amendment analysis. Cf. Spallone v. United States, — U.S. -, 110 S.Ct. 625, 646 n. 12, 107 L.Ed.2d 644 (1990) (Brennan, J., joined by Marshall, Blackmun, and Stevens, JJ., dissenting) (characterizing as “unpersuasive” the argument that legislative voting is speech protected by the First Amendment; “[wjhile the act of publicly voting on legislation arguably contains a communicative element, the act is quintessential^ one of governance”; the majority did not reach the issue.); Clarke v. United States, 886 F.2d 404, 418 (D.C.Cir.1989) (Buckley, J., concurring) (describing issue as “an important case of first impression”). Its shadow would fall on all the myriad conditioned funding provisions used by Congress (see, e.g., South Dakota v. Dole, 483 U.S. 203, 107 S.Ct. 2793, 97 L.Ed.2d 171 (1987)) and by state or local governments.
Vacatur appears particularly appropriate where retention of the precedent creates a gratuitous conflict with a co-equal branch of government. In Matter of City of El Paso, Texas, 887 F.2d 1103, 1106 (D.C.Cir.1989), though by no means suggesting that vacatur for mootness was discretionary, we emphasized the avoidance of constitutional questions as one of the bases for vacatur. Similarly, in a context where the Supreme Court has explicitly regarded its application of the mootness doctrine as discretionary— where the mootness applies only to named members of a plaintiff class — it has viewed the constitutional character of the decision as compelling vacatur. Kremens v. Bartley, 431 U.S. 119, 133-34 & n. 15, 97 S.Ct. 1709, 1717 & n. 15, 52 L.Ed.2d 184 (1977) (citing Justice Brandeis’s concurring opinion in Ashwander v. TVA, 297 U.S. 288, 341, 56 S.Ct. 466, 480, 80 L.Ed. 688 (1936)). The point is equally applicable here.
Appellees suggest that the government here is in some kind of logical bind: if the issue is not “capable of repetition yet evading review,” they suggest, then preserving the decision can have no impact. Of course the argument ignores the core of the adjudicatory process, reasoning by analogy. As Judge Buckley noted in his concurrence, “we may have opened the door to more litigation than we can now appreciate.” Clarke, 886 F.2d at 418. It being unnecessary to open the door at all on the present facts, we return it to its pre-existing condition, neither open nor shut. The decision of the panel is vacated, and case is remanded to the district court to vacate its decision and to dismiss Count III of the complaint as moot.
So ordered.
. Appellees have not pointed to any funds covered by § 108’s reference to express exceptions that remained unexpended at the time the 1990 Act became law; it seems probable that if any existed the 1990 Act superseded the authority of the 1989 one.
. Justices Marshall and Brennan concluded that the case was moot, because the federal decision would defeat any attempted state prosecution. Id. at 655-64, 102 S.Ct. at 2647-52 (Marshall, J., dissenting). While they stated that federal courts had the power to issue a more limited injunction, one “subject to the condition that if *702the statute is later found to be valid, the State is free to seek penalties for violations that occurred during the period the injunction was in effect,” id. at 656, 102 S.Ct. at 2648, they asserted a power in the federal courts to issue injunctions providing permanent protection, and would have presumed that any ambiguous injunction did so, id. at 657-58, 102 S.Ct. at 2648-49.
. But cf. Burke v. Barnes, 479 U.S. 361, 364, 107 S.Ct. 734, 736-37, 93 L.Ed.2d 732 (1987) (rejecting argument that possibility of action to recover funds allegedly spent in violation of Anti-Deficiency Act saves a case from being moot, stating that ”[t]here is no indication of a presently existing dispute as to the accounting obligations, and if such a dispute were to arise it would not be between the parties to this case.”).
. It may be that the requirement of probable recurrence between the same parties is dropped where it is likely that the issue will recur between the defendant and others without ever reaching the Supreme Court. See Honig v. Doe, 484 U.S. 305, 335-36, 108 S.Ct. 592, 610, 98 L.Ed.2d 686 (1988) (Scalia, J., dissenting).
. Thus the cases talk variously of either the issuance of the mandate, or of the filing (or disposition) of an application for rehearing, as the moment that brings to an end the duty to vacate a decision for prior mootness. Of course normally these run close together; under Rule 41(a) of the Federal Rules of Appellate Procedure the filing of a petition for rehearing stays the mandate automatically.