with whom Circuit Judge T.G. NELSON joins, dissenting.
I
I agree with this much of the majority’s claim that there are exceptional eireum-*1067stances in this case: Not every day does a court of appeals recall the mandate in order to reconsider an argument that has already been presented to — and rejected by — the Supreme Court in a petition for a writ of certio-rari. So far as I know, this is a first anywhere, anytime. What happened here, then, to justify such an extraordinary departure from orderly procedure? The majority suggests that “our normal en banc process did not function in the intended manner.” Maj. op. at 1048. In fact, nothing at all unusual happened; the process operated just as it’s supposed to.
Here is how it all works: All suggestions for rehearing en banc, whether or not bundled with a petition for panel rehearing, are first considered by the panel that issued the opinion. See Circuit Advisory Committee Note (4) to Rules 35-1 to 35-3. A judge who has doubts about the panel’s opinion must await the panel’s action before calling for a vote on whether to take the case en banc, since the panel might solve the problem by amending the opinion. In order to be notified when the panel has acted, an off-panel judge must ask the panel for a 5.4(b) notice, named after our General Order of that number. The 5.4(b) notice, when issued by the panel, informs the full court of the panel’s action and triggers the time for any judge still dissatisfied with the opinion to make an en banc call; the call must be made within 14 days of the issuance of the 5.4(b) notice. See G.O. 5.4(b)(2). These are the procedures applicable to all cases and every one of our judges is familiar with them; we use them every day.
In this case, an off-panel judge (Judge X) made a request for a 5.4(b) notice on September 27, 1996, and the panel duly issued that notice on January 17, 1997.1 In that notice, the panel informed the court that it was planning to make some minor changes to the opinion and otherwise to deny the petition. That memo was circulated, as is customary, by way of our internal e-mail system. Again, under our General Orders, if no judge calls for an en banc vote within 14 days of the 5.4(b) notice, the suggestion for rehearing en banc is deemed rejected and the panel is free to file an order to that effect. Adding 14 days to January 17 brings us to January 31; no en bane call was made as of that date. In an abundance of caution, the author of the opinion inquired of the judge charged with coordinating the en banc process and was advised that January 31 was, indeed, the last date to make an en banc call. Over a month later, on March 6, the order denying rehearing was filed with a revised opinion containing the small amendments mentioned by the panel in its 5.4(b) notice.
Six days later, on March 12, another off-panel judge (Judge Y) wrote to the panel asking that it withdraw the order denying rehearing in order to give that judge an opportunity to make an en banc call. The judge in question explained as follows the failure to make a timely call:
I ... attempted to determine why I had not become aware of your decision earlier. The answer appears to be that my chambers systems malfunctioned and the opinion simply fell between the cracks. A partial explanation, but not excuse, is that the disposition was circulated shortly before a law clerk transition and that the old and new law clerks assigned to the case failed to communicate.
The following day, March 13, Judge X, who had originally requested the 5.4(b) notice, seconded the request. Four days after that, Judge X circulated a second memo noting that a timely 5.4(b) request had been made. “Was a 5.4(b) notice circulated? Did I miss it?”, inquired Judge X.
The author of the opinion responded, noting the mandate had been stayed to allow petitioner to seek certiorari. On the question whether an en banc call could be made, the authoring judge stated as follows:
*1068On Monday, March 17, 1997, I sent you copies of papers from my file that demonstrate that you both had adequate notice of the panel’s intent to amend the opinion and a 5.4(b) notice. The notice was sent to Associates on January 17,1997.
The memo also noted that “[i]n compliance with our general orders, the en banc coordinator set the date of January 31, 1997 as the last date on which a timely call for en banc could be made” and confirmed that “the panel was under no further obligation to do anything in response to [the] request for a 5.4(b) notice dated September 27,1996.”
Neither Judge X nor Judge Y responded by claiming they did not receive the 5.4(b) notice. Neither argued at the time that they should be allowed to make a belated en banc call because of excusable neglect.2 Neither judge notified the full court that he felt aggrieved by the panel’s unwillingness to waive the deadline as a matter of comity. In fact, this whole controversy — the claim that there was some compromise of the court’s processes — did not come to the attention of the full court until after the Supreme Court denied certiorari, even though issuance of the mandate was safely stayed that entire time and any adjustment could have been accomplished without the extraordinary step of recalling the mandate.3
This recitation of events should make clear that there was nothing unusual about what happened in this ease. The panel followed our procedures scrupulously; the en banc coordinator made an unremarkable ruling; there was no breakdown of our internal communications system. Two judges made mistakes — precisely the sort of mistakes lawyers make all the time and as to which we routinely hold they do not amount to excusable neglect. See, e.g., Committee for Idaho’s High Desert, Inc. v. Yost, 92 F.3d 814, 824 (9th Cir.1996), (“inadvertence, ignorance of the rules, or mistakes construing the rules do not usually constitute ‘excusable’ neglect,” quoting Pioneer Inv. Servs. Co. v. Brunswick Assocs. Ltd. Partnership, 507 U.S. 380, 392, 113 S.Ct. 1489, 1495, 123 L.Ed.2d 74 (1993)). At the same time, 35 other judges who were entitled to make an en banc call had reviewed the slip opinion and the suggestion for en banc rehearing and were satisfied that the result did not meet the rigorous stan*1069dards for en bane review. Small wonder. While the majority now comes to a different conclusion than the three-judge panel, it does not indicate where the panel opinion was in conflict with the law of the circuit, the law of another circuit or the Supreme Court. In fact, the difference between the earlier opinion and this one involves only the application of law to facts, quintessentially the kind of case where en banc review is routinely rejected. Nor is this a case of transcending importance, unless one views every death case as exceptional. While the panel might have been more accommodating to Judges X and Y, the action is not unprecedented, particularly when the request was presented six weeks late and without any compelling reason for the delay.4
The majority also seems to suggest that Judges X and Y were misled because they were told they could not call for en banc, while the filing of an amended opinion in fact re-opened the period for making an en banc call. Maj. op. at 1049 & n. 2. The short answer is that they were not misled. As the memo from the authoring judge quoted above makes clear, the en bane coordinator’s ruling only held that the panel was “under no further obligation to do anything in response to Judge [X]’s request for a 5.4(b) notice dated September 27, 1997.” Nothing and no one dissuaded Judges X or Y, or both together, from making a sua sponte en banc call on the basis of the amended opinion. No one addressed the point at all.
But maybe Judges X and/or Y were not aware of the possibility they could call for en banc review on the basis of the amended opinion, and were misled by the failure of the panel and the en banc coordinator to offer them this alternative. Not so. Judge X in the memo of March 13, 1997, said as follows: “I wonder whether in any event an amended opinion triggers a new 5.4(b) period.” Since this was written only 7 days after the amended opinion had been filed, nothing prevented Judge X, Judge Y or any other judge (we were all copied in on the memo) from seizing this thought and asking for a 5.4(b) notice right there and then. The judge in question would not even have had to be sure that it would work; the request could have been made and the matter “litigated” before the en banc coordinator and the full court. See n. 2 supra. Once again, the Judges Appellant had a full and complete remedy for the supposed breakdown in our processes, but did nothing.
Where then is the beef? Where the extraordinary circumstances that empower us to exercise the seldom-used authority to recall the mandate? Where the affront to “the integrity of our processes” the majority complains about? Maj. op. at 1048. It just doesn’t exist.
But let’s assume, contrary to fact, that Judges X and Y were somehow hoodwinked out of their right to make an en banc call. It does not follow that this is something the petitioner can complain about. It has long been established that “[t]he function of en banc hearings is not to review alleged errors for the benefit of losing litigants,” United States v. Rosciano, 499 F.2d 173, 174 (7th Cir.1974) (en banc), citing Western Pacific R.R. Corp. v. Western Pacific R.R. Co., 345 U.S. 247, 256-59, 73 S.Ct. 656, 661-62, 97 L.Ed. 986 (1953). En banc review, rather, is a means for a multi-judge court that sits in panels of less than the entire number to bring consistency within its law. See Western Pacific R.R., 345 U.S. at 260 n. 20, 73 S.Ct. at 663 n. 20. Thus, the majority commits egregious error when it suggests that en banc review is “a part of the full and fair procedure our court affords in connection with initial habeas petitions.” Maj. op. at 1049. A cursory review of the most elemental authorities on this point discloses this is just not so. The statute which grants power to the court to sit en banc gives the parties no rights whatsoever to require en bane rehearing, or even a consideration by the full court of the suggestion. See, e.g., Western Pacific R.R., passim; Notes of Advisory Committee on Appellate Rules following Fed. R. App. P. 35 (“The provision that a vote will not be taken as a result of the suggestion of the *1070party unless requested by a judge of the court in regular active service or by a judge who was a member of the panel that rendered a decision sought to be reheard is intended to make it clear that a suggestion of a party as such does not require any action by the court.”).
The implications of this body of law for orn-ease are rather clear. Since the en banc process gives no rights to the parties, a breakdown in the process — even if there was one — creates no problem of extraordinary dimensions. If the en banc call is missed for whatever reason, the error can be corrected in a future case where the problem again manifests itself. Any problem worthy of en banc consideration will perforce appear again and again; missing an en banc call in a single case does not terminate forever the opportunity of judges troubled by the error to raise the issue. Thus, even were I to accept the majority’s premises that something untoward happened here, I cannot see how this supports the extraordinary remedy of recalling the mandate.
That this is a capital case does not change the calculus. The stakes are higher in a death ease, to be sure, but the stakes for a particular litigant play no legitimate role in the en banc process. Nor can we say that “death is different” and therefore that it is appropriate to impose more layers of procedural safeguards. The Supreme Court has already imposed quite a few layers of additional safeguards with respect to death cases. There is no indication anywhere in the Court’s opinions that we are free to add to these safeguards by imposing an especially rigorous en banc review process. Indeed given that 28 U.S.C. § 46(c) and the 1967 Federal Rules of Appellate Procedure Amendments all were promulgated prior to Furman v. Georgia, 408 U.S. 238, 92 S.Ct. 2726, 33 L.Ed.2d 346 (1972), it would seem quite strange to extract from them special procedures peculiarly applicable to death cases.
Ultimately, then, the majority’s decision to recall the mandate must stand or fall on the other factors it cites in support of its action, such as Thompson’s claim of actual innocence, the supposed errors in the panel’s decision and the serious consequences to the petitioner. As Judge Hall points out, there are few petitioners who can’t make out a colorable claim that some or all of these factors are present, and we will be hard pressed to distinguish them from Thompson’s ease. See Hall dissent at 1065-66. Given the difficulties imposed in filing second habeas petitions under AEDPA, we can expect a flood of requests to “recall the mandate”.
II
While I would not reach the merits of the first habeas petition, I comment briefly on the portion of Judge Fletcher’s opinion dealing with prosecutorial misconduct. Fletcher op. Part III. Because that portion of the opinion does not command a majority, it is more in the nature of ruminations by some of our judges. See also Tashima concurrence. As I find this issue troubling, I contribute a few random thoughts of my own.
To begin with, I do not agree with Judge Fletcher’s broad statement that “it is well established that when no new significant evidence comes to light a prosecutor cannot, in order to convict two defendants at separate trials, offer inconsistent theories and facts regarding the same crime.” Fletcher op. at 1058. There is, in fact, a long line of cases that says, if only by way of dicta, that judicial estoppel will not apply against the government in criminal cases. See, e.g., Nichols v. Scott, 69 F.3d 1255, 1272 (5th Cir.1995); United States v. McCaskey, 9 F.3d 368, 378 (5th Cir.1993); United States v. Kattar, 840 F.2d 118, 129-30 n. 7 (1st Cir.1988) (all noting that judicial estoppel has never been applied against the government in a criminal case). The matter is thus far from settled, much less “well established.”
That said, there is surely something troubling about having the same sovereign, particularly acting through the same prosecutor, urge upon two juries a conviction of both A and B, when it is clear that the crime was committed by either A or B. To begin with, it raises the suspicion that the prosecutor may have presented testimony he knows, or has reason to believe, is false. If that be the case, the breach in prosecutorial ethics con*1071sists of putting on the tainted testimony, not in pursuing the inconsistent verdicts. But it is impossible to make judgments about what the prosecutor knew or should have known at our level, as Judge Tashima points out. See Tashima concurrence at 1064; but cf. Fletcher op. at 1056 (finding prosecutor’s characterization of testimony “patently untrue”). Thus, if the petitioner makes a prima facie case that the prosecutor knowingly presented false evidence, the matter must be resolved at an evidentiary hearing.
But, as Judge Kleinfeld points out, prosecutors are not omniscient. See Kleinfeld dissent at 1074-75. They may be confronted with witnesses who present mutually inconsistent versions of what happened, and there may be no way of knowing which version — if any — is true. Is the prosecutor then precluded from presenting either case to the jury? Must he pick one based on his intuition? I believe not. A prosecutor, like any other lawyer, is entitled to retain skepticism about the evidence he presents and trust the jury to make the right judgment. After all, the guarantee of due process encompasses a fair trial before a fair judge and jury; the right to a lawyer and to exculpatory evidence available to the prosecution; and the right not to have the prosecutor lie to the jury. But I cannot see that it encompasses the right to have a prosecutor who is convinced of the defendant’s guilt. We trust the adversary process, the good sense of jurors, the presumption of innocence and the prosecution’s heavy burden of proof to ensure a verdict that is fair to the defendant. If the system works as it should, A and B both may be acquitted, but in no event should more than one of them be convicted.
Must we be troubled, however, where the process does not work well — where both A and B get convicted, despite the best efforts of everyone involved? I believe so. If A and B cannot both be guilty, we then know that one innocent person has been convicted. Whether this rises to the level of a due process violation is close to the question the Supreme Court posed in Herrera v. Collins, 506 U.S. 390, 113 S.Ct. 853, 122 L.Ed.2d 203 (1993): Is there a federal constitutional right not just to fair process, but also to a correct result? While the Supreme Court has not yet answered this question, the better view seems to be that the state has no right to execute an innocent man, however fairly it has obtained the conviction. In the case of mutually inconsistent verdicts, which I am not sure is the case here, I believe that the state is required to take the necessary steps to set aside or modify at least one of the verdicts. While the authority for this position is far from ironclad, it is nonetheless compelling.
I start with Durley v. Mayo, 351 U.S. 277, 76 S.Ct. 806, 100 L.Ed. 1178 (1956), where the Court granted certiorari to determine whether due process was offended by a conviction which was alleged to rest on perjured testimony, but the prosecutor had not been aware of the perjury at trial. The Court dismissed for lack of jurisdiction, but four Justices (Douglas, J., joined by Warren, C.J. and Black and Clark, JJ.) would have held that it is a denial of due process for the state to retain a conviction it later discovers was based on perjury. The Second Circuit in Sanders v. Sullivan, 863 F.2d 218 (2d Cir. 1988) (Kaufman, J.) took the same view.
Writing for himself and Justice Ginsburg, Justice Stevens relied on Durley in arguing that a stay should be granted in Jacobs v. Scott, 513 U.S. 1067, 115 S.Ct. 711, 130 L.Ed.2d 618 (1995). (Justice Breyer would also have issued a stay.) Jacobs had been convicted of capital murder, and the state had then tried and convicted Jacobs’s sister, based on Jacobs’s own testimony. At the sister’s trial, the prosecutor had argued that he now believed Jacobs had told the truth and it was the sister who killed the victim. Nevertheless, Texas insisted on executing Jacobs, and eventually did. Justice Stevens’s dissent argues, like Judge Fletcher here, that the prosecutor committed misconduct at trial in presenting inconsistent theories. The facts in Jacobs are more compelling than in our case, but they raise the same basic issues. Whether or not one agrees with Justice Stevens and the judges joining Part III of Judge Fletcher’s opinion that the prosecutor committed misconduct at trial by pressing inconsistent theories, it still seems mighty troubling for the state to take a *1072prisoner’s life after having publicly announced that it believes him to be innocent. I, however, would treat it as a denial of substantive due process rather than prosecu-torial misconduct.
Which brings me to a final thought: Whatever the Supreme Court ultimately decides due process requires, it is unseemly for the state to act as it did in Jacobs and, perhaps, in our case as well. Whether or not the United States Constitution allows them to argue inconsistent theories to different juries, it surely does not inspire public confidence in our criminal justice system for prosecutors to leave themselves open to charges of manipulation. The danger is particularly grave in capital cases, where the manipulation could well cause the execution of an innocent person. That is a calamity none of us should contemplate lightly.
. I regret having to delve into internal court communications, especially as I do not believe any of this matters at all. See pp. 1069-70 infra. However, as the majority has opened the door by giving what I see as an inaccurate description of what happened, I feel constrained to set the record straight. I believe this is especially appropriate since the parties do not have access to these facts and thus have no way of challenging or defending the majority's assertions.
. While our General Orders provide time limits for taking certain actions, including the making of an en banc call, they are not self-executing. A judge who disagrees with the application of the General Orders can challenge the action and, ultimately, put the matter to a vote of the whole court. In this case, nothing precluded Judges X or Y from calling for en banc late. Had such a call been made, it would have been up to the en banc coordinator to determine whether it was timely; the en banc coordinator has traditionally exercised authority to waive certain time limits for excusable neglect and the like. Had the en banc coordinator ruled the call untimely, the aggrieved judges would have had two remedies: They could have appealed the ruling to the full court, arguing for a waiver on the basis of excusable neglect or collegiality or whatever, or they could have asked for a suspension of the rules pursuant to G.O. 11.11. The latter might have been harder than the former, but they could have done either or both. They did neither.
. Judge Reinhardt purports to dispute my account of what transpired. He is mistaken. As I state in the text, the full court was notified in March that Judges X and Y had missed the deadline. See p. 1069 infra (“Since [Judge X's memo] was written only 7 days after the amended opinion was filed, nothing prevented Judge X, Judge Y or any other judge (we were all copied in on the memo)____") (emphasis added). What no one knew until July is that Judges X or Y thought this reflected some "procedural misunderstandings” or that "our en banc process did not function in the intended manner.” Maj. op. at 1048. Indeed, in his memo of March 12, Judge Y takes the entirely contrary position: "At this point all I can do is ask whether the panel might be willing to recall the mandate to permit me to make a prompt though belated call. Under our rules, the decision is clearly yours." (emphasis added).
Judge Reinhardt also seems to dispute my assertion that the full court was not notified of the controversy until after the Supreme court denied certiorari. He says: "In addition, before our mandate was spread in the district court, two other judges also requested that a belated en banc call be entertained. Again, the entire court was notified immediately.” Reinhardt concurrence at p. 1060. Here is what happened:
June 2, 1997 Supreme Court denies certiorari.
June 11, 1997 Ninth Circuit issues mandate.
July 7, 1997 Judge Z calls for belated en banc. Judge Z's memo of July 7, almost a month after the mandate was issued, was the first that the court at large became aware that "there were procedural misunderstandings” or that "our en banc process did not function in the intended manner.” Judge Reinhardt's assertion that "the entire court was notified immediately” is contrary to fact.
. The majority’s assertion that “there has been no occasion in the past several decades in which this court has made a similar error in any type of case,” maj. op. at 1050, is not accurate. It’s happened to me more than once.