dissenting:
The defendants applied to the district court to continue bail pending their appeals from their convictions and sentences for mail and wire fraud and related offenses. The district court granted their applications, and the defendants were released after posting bonds of $1 million (O’Keefe), $500,000 (Schmidt), $500,000 (O’Brien) and $250,000 (Bennett). The government appealed from the district court’s order as authorized by 18 U.S.C. § 3731.
Rule 9(b) of the Federal Rules of Appellate Procedure provides that “[ajpplication for release after a judgment of conviction shall be made in the first instance in the district court.” In reviewing the district courts decision, a court of appeals is free to make an independent determination on the merits of the prisoner’s application. United States v. Clark, 917 F.2d 177, 179-80 (5th Cir.1990); United States v. Hawkins, 617 F.2d 59 (5th Cir.), cert. denied, 449 U.S. 952, 101 S.Ct. 355, 66 L.Ed.2d 215 (1980); United States v. Provenzano, 605 F.2d 85, 92-93 (3rd Cir.1979). However, the district court’s conclusion is entitled to “great deference.” Harris v. United States, 404 U.S. 1232, 1232, 92 S.Ct. 10, 30 L.Ed.2d 25 (1971); United States v. Oliver, 683 F.2d 224, 235 (7th Cir.1982); United States v. Gigax, 605 F.2d 507 (10th Cir.1979); United States v. Provenzano, 605 F.2d at 91-92. See United States v. Crabtree, 754 F.2d 1200 (5th Cir.1985).
To obtain release pending appeal, a convicted defendant must establish four factors: (1) that he is not likely to flee or pose a danger to the safety of others; (2) that the appeal is not for purpose of delay; (3) that the appeal raises a substantial question of law or fact; and (4) that the substantial question, if decided favorably to the defendant, is likely to result in reversal, in an order for a new trial, in a sentence without punishment, or in a sentence with reduced imprisonment. 18 U.S.C. § 3143(b). United States v. Clark, 917 F.2d at 179. In its order admitting the defendants to bail pending appeal, the district court found that the defendants passed all four prongs of this test. Reviewing the district court’s written reasons and oral statements at the hearing on this issue with deference, and independently reviewing relevant portions of the record of the trial and post-trial proceedings, I concur with the district court’s determinations, and would affirm the district court’s judgment admitting the defendants to bail pending their appeals.
In the present case, the only prong of the four-part test that is problematic or that warrants any discussion is the third one: whether the defendants’ appeals raise a substantial question of law or fact, i.e., “ ‘one of *282more substance than would be necessary to a finding that it was not frivolous[;] ... a “close” question or one that very well could be decided the other way.’ ” United States v. Valera-Elizondo, 761 F.2d 1020, 1024 (5th Cir.1985)(quoting United States v. Giancola, 754 F.2d 898, 901 (11th Cir.1985)).
I.
After the jury convicted the defendants of conspiracy, wire fraud, mail fraud, and money laundering, the trial judge, in the same order in which he recused himself, granted the defendants’ motion for a new trial; several weeks later, he denied the government’s motion for reconsideration. The government filed an interlocutory appeal under 18 U.S.C. § 3731 contesting the trial judge’s new trial order and contending that the trial judge’s order denying the government’s motion for reconsideration was void because of his prior order disqualifying himself in the case. In O’Keefe I, a panel of this court held that the district court judge erred in performing a discretionary act by ruling on the motion for reconsideration after he had recused himself, but that the error did not have to be vacated because it was “harmless.”1 O’Keefe I, 128 F.3d at 891, 892-93. The O’Keefe I panel then proceeded to review the district judge’s ruling on the defendants’ motion for new trial on the merits. The panel concluded that the judge’s ruling constituted an abuse of discretion or legal error in that (i) the government’s knowing failure to correct perjured testimony did not violate the defendants’ due process rights under Napue v. Illinois, 360 U.S. 264, 79 S.Ct. 1173, 3 L.Ed.2d 1217 (1959), because the falsehoods were not ma-ferial, i.e., there was not “a reasonable probability that the jury would have reached a different outcome even had it been fully aware of all of the alleged inconsistencies and falsehoods in [the prosecution witness] Donaldson’s testimony,” O’Keefe I, 128 F.3d at 898; (ii) in the absence of a material Napue violation, the trial judge’s additional findings, viz., that the government impermissibly delayed the disclosure of Brady material, that the inconsistencies in prosecution witness Moore’s testimony clouded or weakened the government’s case, and that the prosecution attempted to mislead the defense by changing the indictment, were insufficient to warrant the granting of a new trial in the “interests of justice” under Fed.R.Crim.P. 33.2
II.
The government contends that the defendants’ appeal cannot raise any “substantial issue of law” with respect to government misconduct and perjury by prosecution witnesses because any such purported issue is foreclosed by the law of the case doctrine. The government argues that United States v. O’Keefe, 128 F.3d 885 (5th Cir.1997) (O’Keefe I), decided upon rules of law that will continue to govern the same issues during the defendants’ appeal of right from final judgment. However, whether the law of the case doctrine precludes the direct appeal panel from considering such issues is itself a threshold “substantial issue of law.” The question of the applicability of the law of the case doctrine is substantial, not only because it is “close” and of more substance than “non-frivolous” questions, but also because it in*283volves novel issues concerning the recusal of judges, harmless- error, and the effects of government interlocutory appeals in criminal cases upon defendants’ fundamental rights to appeal and to have a full and fair day in court.
As defined by the Supreme Court, the doctrine of the law of the case “ ‘posits that when a court decides upon a rule of law, that decision should continue to govern the same issues in subsequent stages of the same case.’ This rule of practice promotes the finality and efficiency of the judicial process by ‘protecting against the agitation of settled issues.’ ” Christianson v. Colt Indus. Operating Corp., 486 U.S. 800, 816, 108 S.Ct. 2166, 100 L.Ed.2d 811 (1988) (citing Arizona v. California, 460 U.S. 605, 618, 103 S.Ct. 1382, 75 L.Ed.2d 318 (1983) and 1B J. Moore, J. Lucas, & T. Currier, Moore’s Federal Practice ¶ 0.404[1], at 118 (2d ed.1984)).
The law of the case doctrine attaches in interlocutory appeals only upon matters that have actually been decided. As to decisions upon rules of law, the interlocutory appeal establishes the law of the case. Royal Ins. Co. v. Quinn-L Capital Corp., 3 F.3d 877, 881 (5th Cir.1993). Factual determinations in an interlocutory appeal will generally not establish the law of the case. Id. See 18 Moore’s Federal Practice § 134.20 (3d ed.1998) (“Unlike the doctrine of claim preclusion, the law of the case doctrine does not apply to issues or claims that were not actually decided; for this reason, failure to raise an issue on interlocutory appeal should not operate to preclude the issue on a later appeal from a final judgment, even if other issues were raised by the party or an opponent in a permitted interlocutory appeal.”).
The law of the case doctrine applies to an issue that has actually been decided, not to statements made by the court in passing, or stated as possible alternatives, or dictum. 18 Moore’s Federal Practice §§ 134.20[3], 134.21[2] (3d. ed.1998) (citing, e.g., Royal Ins. Co. v. Quinn-L Capital Corp., 3 F.3d at 880; Great Lakes Dredge & Dock Co. v. Tanker Robert Watt Miller, 957 F.2d 1575, 1578 (11th Cir.), cert. denied, 506 U.S. 981, 113 S.Ct. 484, 121 L.Ed.2d 388 (1992)).
When the law of the case doctrine is applied by a court to its own prior decisions, it is properly characterized as discretionary in nature. 18 Moore’s Federal Practice § 134.21[1] (3d ed.1998). The doctrine “ ‘merely expresses the practice of courts generally to refuse to reopen what has been decided, not a limit to their power.’ ” Christianson v. Colt Industries, 486 U.S. at 817, 108 S.Ct. 2166 (quoting Messinger v. Anderson, 225 U.S. 436, 444, 32 S.Ct. 739, 56 L.Ed. 1152 (1912)(Holmes, J.) (citations omitted)). “A court has the power to revisit prior decisions of its own or of a coordinate court in any circumstance, although as a rule courts should be loathe to do so in the absence of extraordinary circumstances such as where the initial decision was ‘clearly erroneous and would work a manifest injustice.’ ” Id. (quoting Arizona v. California, 460 U.S. at 618, n. 8,103 S.Ct. 1382 (citation omitted)).
In this Circuit, we have described the nature of the law of the case doctrine and its exceptions in similar fashion:
While application of the doctrine is discretionary, this court will generally refuse to revisit a prior panel’s decision unless “(i) the evidence on a subsequent trial - was substantially different, (ii) controlling authority has since made a contrary decision of the law applicable to such issues, or (iii) the decision was clearly erroneous and would work a manifest injustice.”
Free v. Abbott Laboratories, 164 F.3d 270, 272-73 (5th Cir.1999) (quoting North Mississippi Comms., Inc. v. Jones, 951 F.2d 652, 656 (5th Cir.1992)). See United States v. Becerra, 155 F.3d 740, 752-753 (5th Cir.1998).
Applying these principles, it is evident that the defendants’ appeal and the government’s assertion of the law of the case bar raise substantial questions of law with respect to whether the O’Keefe I panel committed clear error that will work manifest injustice by (a) holding that a trial judge’s legal error in knowingly performing a discretionary judicial act in violation of his own order disqualifying himself under 28 U.S.C. § 455(a) can be “harmless error” that does not have to be vacated; and (b) failing to hold that both of the trial judge’s rulings, i.e., his grant of the *284defendant’s new trial motion and his denial of the government’s motion for reconsideration, were discretionary acts performed in violation of his disqualification order, that the rulings must be vacated, and that the case must be remanded for further proceedings before a different judge.
(a)
The trial judge disqualified himself in accordance with 28 U.S.C. § 455(a), which provides that “[a]ny justice, judge, or magistrate ' of the United States shall disqualify himself in any proceeding in which his impartiality might reasonably be questioned.” The O’Keefe I panel correctly held that “[ojnce a judge recuses himself from a case, the judge may take no action other than the ministerial acts necessary to transfer the case to another judge, even when recusal is improvidently decided.” O’Keefe I, 128 F.3d at 891 (citing “Doddy v. Oxy USA Inc., 101 F.3d 448, 457 (5th Cir.1996)(holding that judge erred in vacating recusal order after recusing herself); Moody v. Simmons, 858 F.2d 137, 143 (3rd Cir.1988)(stating that judge may only perform the ‘housekeeping’ duties necessary to transfer a case to another judge after recusing himself from a proceeding)”).
There is a substantial question, however, whether O’Keefe I clearly erred in holding that a trial judge’s discretionary rulings in a criminal case in violation of his own order of disqualification can be harmless and may not require his infringing orders to be vacated. Before O’Keefe I, this court and other federal courts of appeals had held consistently in both civil and criminal cases that such an error requires the appellate court to vacate the offending discretionary order and to remand the case for reassignment to a different judge.3 See, e.g., Moody v. Simmons, 858 F.2d 137,143 (3d Cir.1988)(“Once a judge has disqualified himself, he or she may enter no further orders in the case. His power is limited to performing ministerial duties. * * * A judge who was obliged to recuse acts outside his jurisdiction [or] commits a clear error of law.... Mandamus is thus the proper remedy to vacate the orders of a judge who acted when he should have recused.”); Stringer v. United States, 233 F.2d 947, 948 (9th Cir.1956)(“[O]nce having disqualified himself for cause, on his own motion, it was incurable error for the district judge to resume full control and try the case.”); United States v. Feldman, 983 F.2d 144 (9th Cir.1992)(A judge cannot order that his recusal from a proceeding be limited to certain aspects or issues of the case. An order denying complete recusal must be vacated and the case reassigned to a different judge.); McCuin v. Texas Power and Light Co., 714 F.2d 1255, 1260-61 (5th Cir.1983)(A recused judge’s order reassigning case to a particular judge must be vacated and the case remanded according to local practice, or *285in the absence thereof, by the senior active judge.); El Fenix de Puerto Rico v. The M/Y JOHANNY, 36 F.3d 136, 141 (1st Cir.1994)(“As a general rule, a trial judge who has recused himself should take no other action in the ease except the necessary ministerial acts to have the case transferred to another judge____ [A] recused judge’s power is limited to performing [such] ministerial duties ...” (internal quotation marks and citations omitted)). In fact, in O’Keefe I the government relied on these same authorities in urging the appellate panel to vacate the trial judge’s denial of the motion for reconsideration as void for having been entered after the judge disqualified himself. Brief for Appellant, United States v. O’Keefe, 128 F.3d at 885. Consequently, it is at least a “close” question of “more substance than would be necessary to a finding that it was not frivolous,” whether O’Keefe I clearly erred in departing from the precedents of this and all other federal circuits which hold that a trial judge’s deliberate performance of a discretionary judicial act in violation of his own order disqualifying himself must be vacated.
There is also a substantial question as to whether O’Keefe I clearly erred in concluding that Liljeberg v. Health Services Acquisition Corp., 486 U.S. 847, 108 S.Ct. 2194, 100 L.Ed.2d 855 (1988) created a new, special harmless error test for determining whether “any order that a judge issues after the judge has, or should have, recused himself must be vacated.” O’Keefe I, 128 F.3d at 892. It may be cogently argued that Lilje-berg does not create a special harmless error test at all; does not expressly or impliedly require, by calling attention to three appropriate equitable considerations for courts to use in determining whether a party should be relieved of a final civil judgment under Fed.R.Civ.P. 60(b)(6) based on a judge’s § 455(a) violation, that the same considerations be applied outside of the Rule 60(b) motion context; and certainly does not require or contemplate that Rule 60(b)(6), or those three equitable considerations, be applied in criminal cases.
In Liljeberg, a party in a civil case filed a motion under Fed.R.Civ.P. 60(b)(6) to be relieved from a final civil judgment on the ground that the trial judge, during the trial and rendition of the judgment, was, unbeknownst to the movant, a fiduciary of a university having a substantial financial interest in the outcome of the case. On the movant’s second appeal to this court, 796 F.2d 796 (5th Cir.1986); see also 747 F.2d 1463 (5th Cir.1984) (unpublished table decision), we granted the motion to vacate the judgment and remanded the case for a new trial or other further proceedings.
On certiorari, the Supreme Court affirmed. Liljeberg v. Health Services Acquisition Corp., 486 U.S. 847, 108 S.Ct. 2194, 100 L.Ed.2d 855 (1988). The Court addressed the violation of § 455 by the trial judge and the relief of the judgment debtor from the judgment under Fed.R.Civ.P. 60(b)(6) in separate parts of its opinion.
First, the Supreme Court held that a violation of 28 U.S.C. § 455(a), which requires a judge to disqualify himself in any proceeding in which his impartiality might reasonably be questioned, occurs when a reasonable person, knowing the relevant facts, would expect that a judge knew of the circumstances creating an appearance of partiality, notwithstanding a finding that the judge, was not actually conscious of those circumstances. Moreover, in a proper case, § 455(a) applies retroactively, as well as prospectively, and requires a judge, upon discovering that he performed a discretionary judicial act under circumstances that would cause an objective observer to question his impartiality, to rectify an oversight and to take steps necessary to maintain public confidence in the impartiality of the judiciary. For example, such a judge may be required to disqualify himself retrospectively and to vacate his discretionary judicial action in violation of § 455(a). Here, because there was ample basis in the record to support the findings of the courts below that an objective observer would have questioned the original trial judge’s impartially, his performance of discretionary judicial acts at that time was a plain violation of the terms of § 455(a), even though his failure to disqualify himself was the product of a temporary lapse of memory. Id. at 859-61, 108 S.Ct. 2194.
*286Second, the Court held that under the circumstances of the case the proper remedy for the § 455(a) violation was to grant the Rule 60(b)(6) motion of the judgment debtor to be relieved of the judgment and granted a new trial. I paraphrase most of the pertinent paragraph of the Court’s opinion. While § 455 itself does not authorize the reopening of closed civil litigation, Fed. R.Civ.P. 60(b) provides a remedy whereby, in an appropriate case, a party may be relieved of a final judgment. In particular, rule 60(b)(6) grants federal courts authority to relieve a party from a final civil judgment “upon such terms as are just.” The Court had “previously noted that it provides courts authority ‘adequate to enable them to vacate judgments whenever such action is appropriate to accomplish justice,’ Klapprott v. United States, 335 U.S. 601, 614-15, 69 S.Ct. 384, 93 L.Ed. 1099 (1949)[eivil immigration case], while also cautioning that it should only be applied in ‘extraordinary circumstances,’ Ackermann v. United States, 340 U.S. 193, 71 S.Ct. 209, 95 L.Ed. 207 (1950)[same].” Liljeberg, 486 U.S. at 863-64, 108 S.Ct. 2194. After setting forth these precepts, the Court concluded that pertinent paragraph as follows:
Rule 60(b)(6) relief is accordingly neither categorically available nor categorically unavailable for all § 455(a) violations. We conclude that in determining whether a judgment should be vacated [under Rule 60(b)(6) ] for a violation of § 155, it is appropriate to consider the risk of injustice to the parties in the particular case, the risk that the denial of relief will produce injustice in other cases, and the risk of undermining the public’s confidence in the judicial process.
Id. at 864, 108 S.Ct. 2194 (emphasis added).
It may be forcefully argued that, when the underlined sentence above is read in the context of the paragraph in which it occurs and of the opinion as a whole, the reader should understand that the sentence is intended to apply in the context of deciding a motion to relieve a party of a civil judgment under Rule 60(b)(6) and that the bracketed words, “under Rule 60(b)(6),” are implicitly included in the sentence. If so, it is evident that underlined sentence merely sets forth some equitable principles that the Court deemed appropriate for consideration in determining whether a § 455(a) violation creates “extraordinary circumstances” warranting the relief of a party from a final civil judgment under Fed.R.Civ.P. 60(b)(6). It is well recognized that a motion for relief from a civil judgment under Rule 60(b) is addressed to the discretion of the court, e.g., Hand v. United States, 441 F.2d 529 (5th Cir.1971)(tax refund case); Simons v. Gorsuch, 715 F.2d 1248 (7th Cir.1983); Clarke v. Burkle, 570 F.2d 824 (8th Cir.1978), and that equitable principles may be taken into account by a court in the exercise of its discretion under Rule 60(b). Bros Inc. v. W.E. Grace Mfg. Co., 320 F.2d 594 (5th Cir.1963); MIF Realty L.P. v. Rochester Assocs., 92 F.3d 752, 756 (8th Cir.1996).4
In reviewing this court of appeals’ decision in Liljeberg to grant the Rule 60(b)(6) motion to vacate the judgment on the basis of the trial judge’s § 455(a) violation, the Supreme Court used several traditional equitable considerations, as well as the three it had said were appropriate, to determine whether relieving the movant from the final civil judgment under Rule 60(b)(6) was the proper remedy for the trial judge’s violation. Of the equitable factors customarily used by courts in deciding 60(b) motions, the Supreme Court considered whether there had been a timely request for relief, whether a showing of special hardship by reason of reliance on the original judgment had been made, or whether the delay in seeking relief was to any extent due to the fault of the Rule 60(b) movant. Liljeberg, 486 U.S. at 868-69, 108 S.Ct. 2194; See 11 Charles Alan Wright, Arthur R. Miller & Mary Kay Kane, Federal Practice and Procedure § 2857, at 256-62 *287(and cases cited therein); § 2866, at 381-92 (and cases cited therein). The Supreme Court did not suggest that there had been an error in the court of appeal’s decision that needed to be reviewed under a harmless error rule. The words “harmless” and “error” do not appear in this part of the opinion. Liljeberg, 486 U.S. at 868-70, 108 S.Ct. 2194. In approving of the court of appeal’s decision, the Supreme Court called it “an eminently sound and wise disposition of this case,” id. at 870, 108 S.Ct. 2194, not one that was free of “harmless error.” Thus, the Court’s opinion as a whole clearly indicates that the three factors it mentioned as “appropriate to consider” were not meant to be used as a freestanding “harmless error” rule, but as a nonexclusive list of equitable considerations, which courts may use along with other established equitable precepts, to guide sound and wise exercise of judicial discretion in deciding Rule 60(b)(6) motions based on § 455 violations.
Consequently, there is at least a “substantial” or “close” question whether O’Keefe I was clearly mistaken in reading the three appropriate considerations mentioned in Lil-jeberg in isolation as creating a new special freestanding “harmless error” rule, not confined to the context of Rule 60(b)(6) motions, but applicable to all civil and criminal cases involving § 455 violations. Such a broad, discretionary “harmless error” rule would be radically different in language, focus and purpose from the harmless error rules of Fed.R.Civ.P. 61 and Fed.R.Crim.P. 52(a).5 Furthermore, there is a substantial question whether O’Keefe I was clearly in error because it applied the equitable factors in the underscored sentence, not as a harmless error rule, but as a talisman giving it the extraordinary power and discretion to skip over affirming or vacating the trial judge’s ruling on the motion to reconsider to review on the merits the trial judge’s granting of a new trial.6
*288Although the Supreme Court’s Liljeberg opinion refers to harmless error in passing at one point, it may be argued that it is highly doubtful that the court intended to establish a new or special harmless error rule for all cases involving § 455 violations. At the end of Part III, in which the Court analyzed § 455 and concluded that the trial judge in Liljeberg plainly violated the statute, the Court added:
A conclusion that a statutory violation occurred does not, however, end our inquiry. As in other areas of the law, there is surely room for harmless error committed by busy judges who inadvertently overlook a disqualifying circumstance. [FN9] There need not be a draconian remedy for every violation of § 455(a). It would be equally wrong, however, to adopt an absolute prohibition against any relief in cases involving forgetful judges.
Liljeberg, 486 U.S. at 862, 108 S.Ct. 2194. In footnote 9 the Court stated that “[ljarge, multidistrict class actions, for example, often present judges with unique difficulties in monitoring any potential interest they may have in litigation.” Id. at 862 n. 9, 108 S.Ct. 2194. In this connection the Court cited Union Carbide Corp. v. U.S. Cutting Service, Inc. 782 F.2d 710, 714 (7th Cir.1986); In re Cement and Concrete Antitrust Litigation, 515 F.Supp. 1076, 1080 (D.Ariz.1981), large class actions in which the trial judges discovered for the first time, well into the litigation, that each of their spouses owned a relatively small amount of securities in a member of the class. These cases, however, hinged upon potential § 455(b)(4)(financial conflict of interests) violations rather than § 455(a)(ap-pearance of partiality) infractions. Thus, the Court’s vague references to “room for error committed by busy judges who inadvertently overlook a disqualifying circumstance,” and the lack of need for “a draconian remedy for every violation of § 455(a),” or the wrong of an “absolute prohibition against any relief in cases involving forgetful judges,” do not appear to formulate a rule at all. Rather, the comments appear to be a precautionary dictum that, in enforcing § 455(a) & (b), inadvertent violations under extenuating circumstances, as opposed to sheer forgetfulness, may sometimes be disregarded as harmless.
The Court’s Part III dictum on harmless error occurs separately and apart from its discussion in Part IV of relieving a party from a final civil judgment under Rule 60(b)(6), in appropriate equitable and extraordinary circumstances, as a remedy for a judge’s § 455(a)violation. Part III makes but one mention of Rule 60(b)(6), and it is revealing. In footnote 9, after observing that large, multidistrict class actions often present judges with unique difficulties in monitoring any potential interest they may have in the litigation, the Court said, “[o]f course, notwithstanding the size or complexity of the litigation, judges remain under a duty to stay informed of any personal or fiduciary financial interest they may have in cases over which they preside. See 28 U.S.C. § 455(c). The complexity of determining the conflict, however, may have a bearing on the Rule 60(b)(6) extraordinary circumstance analysis.” Id. at 862 n. 9, 108 S.Ct. 2194. The Court did not refer to the Rule 60(b)(6) extraordinary circumstances analysis as a harmless error rule. And the Court did not anywhere suggest that the equitable considerations involved in the Rule 60(b)(6) extraordinary circumstances analysis discussed in Part IV of the opinion may be used in other contexts as a harmless rule. The Court does not regard the two conceptions as fungible; and a very strong argument can be made that courts of appeals should not either.
Finally, there is a substantial question as to whether it was clearly erroneous for O’Keefe I to conclude that the Supreme Court in Liljeberg, a civil action, held that Fed.R.Civ.P. 60(b)(6) motions, or the equitable principles appropriate for use in deciding them, may be used to relieve a party of a *289final judgment or order in a criminal case. Federal Rules of Civil Procedure 1 and 81 provide that those rules shall apply to all suits of a civil nature, whether cognizable as cases at law or in equity except those specifically excepted. Federal Rule of Civil Procedure 60(b), therefore, simply does not provide for relief from a-judgment in a criminal case. See United States v. Mosavi, 138 F.3d 1365, 1366 (11th Cir.1998). A criminal conviction can be attacked by motion under 28 USC § 2255, but only for errors of constitutional dimension. See 13A Charles A. Wright, Arthur R. Miller, Edward H. Cooper, Federal Practice and Procedure, § 3550 (1998 Supp.).
(b)
Under § 455(a), grounds for a judge to disqualify himself arise whenever his impartiality might reasonably be questioned; and, even if the judge was not aware of the circumstances creating an appearance of partiality when it occurred, once he realizes that the impropriety existed he is called upon to take steps necessary to maintain public confidence in the impartiality of the judiciary; in a proper case, the judge may be obliged to disqualify himself retroactively and to vacate any orders entered during the time that a reasonable person would harbor doubts about the judge’s impartiality. Liljeberg, 486 U.S. at 860-61, 108 S.Ct. 2194; Health Services Acquisition Corp. v. Liljeberg, 796 F.2d 796, 802 (5th Cir.1986); Moody v. Simmons, 858 F.2d 137 (3d Cir.1988); Hall v. Small Business Administration, 695 F.2d 175, 179 (5th Cir.1983).
In the present case, the trial judge’s 82 page written memorandum and order contains both his order granting defendants’ motion for a new trial and his order recusing himself from the case. The memorandum and order clearly indicates that the trial judge formed the intention to grant the new trial and to recuse himself either during the oral argument on the motion for new trial or at some time before the written order was prepared and filed. The trial judge’s written memorandum and order states: “Although based on the extensive briefs filed by all the parties I was prepared to deny the defendants’ motion for a new trial, following oral argument I am now persuaded that a new trial is the proper remedy under the -circumstances.” Memorandum and Order at 80 (footnote omitted). In its Conclusion, the memorandum and order provides:
The defendants have been successful in obtaining a new trial. As I said before in addressing the perjury of Charles Donaldson, prior to oral argument I was not inclined to grant this remedy. I believed that the jury, which performed its duty so diligently, had been apprised of all relevant information required to reach a verdict. At oral argument it became apparent that such was not the case.
Because, of the sensitive nature of the court’s inquiry concerning conduct of government counsel, the court’s personal participation and questioning of counsel in connection with that inquiry, and the findings of the court resulting from that inquiry, the court feels compelled to recuse itself from further handling of this matter in accordance with 28 U.S.C. § 455.
Accordingly,
IT IS ORDERED that the defendants’ motion for new trial is GRANTED. (Signature omitted).
Id. at 81-82.
Thus, it is evident that the trial judge’s inquiry of a “sensitive nature” into the conduct of the government attorneys at the oral argument on the new trial' motion, the judge’s “personal participation and questioning of counsel” during that inquiry, and the judge’s findings resulting from the inquiry, caused the trial judge to decide that he could no longer maintain impartiality, that he should recuse himself, and that a new trial should be granted.
Consequently, the defendants’ appeal raises a substantial question as to whether O’Keefe I clearly erred in finding or assuming that the trial judge’s decision and order on the new trial motion distinctly preceded the grounds for his recusal under § 455. The judge’s written memorandum and order indicates that he probably decided that he should disqualify himself before the order was prepared in final form or certainly before it was actually signed and filed. Conse*290quently, the possible influence of the judge’s reasons for disqualification upon his decision of the new trial motion is clearly evident. As a practical matter, the trial judge’s new trial and recusal rulings cannot be hermetically separated either temporally or in substance.
Moreover, because of the judge’s recounting of the events and his mental impressions in his written memorandum and order, it would appear to a reasonable person that the judge lost his ability to maintain his impartiality prior to his granting of the new trial, and that the judge realized that his impartiality at the time he granted the new trial might reasonably be questioned. Accordingly, this is a proper ease for the retroactive application of § 455. The trial judge had a duty to make his disqualification retroactive so as to precede his granting of the new trial and to vacate both his new trial and denial of reconsideration order. Accordingly, there is a substantial question whether O’Keefe I clearly erred in not recognizing the pervasiveness of the trial judge’s violations of § 455, in not vacating both his granting of the new trial and his denial of the motion to reconsider, and in not remanding the case to a different judge to consider the defendants’ motion for a new trial anew, shorn of the disqualified judge’s vacated orders.
(c)
Consequently, the defendants’ appeal raises substantial questions as to whether O’Keefe I clearly erred in reaching the merits of the trial judge’s granting of the new trial and in deciding upon the rules of law it adopted in that merits review. Thus, it would clearly cause manifest injustice to the defendants to preclude them, on the basis of a flawed law of the case application, from presenting in their appeal all of their substantial arguments that constitutional errors and defects in the trial affected their substantial rights and the judgment of the jury.
There is a substantial argument that O’Keefe I’s clearly erroneous interlocutory appellate intervention into the merits of this criminal case during a government appeal would, under improper application of a law of the case bar, make for truncated presentation of the defendants’ issues in their own appeal and risk a failure to have a complete appellate determination of the true nature and seriousness of alleged errors based on all of the evidence and a full and fair opportunity for argument by the parties. Consequently, manifest injustice will result if the defendants are denied a full consideration of their constitutional claims and a full vindication of their constitutional rights on direct appeal.
III.
If the direct appeal panel decides that it is not bound by O’Keefe I under the law of the case doctrine, the defendants’ appeal raises substantial questions of law or fact which, are, additionally, so integral to the merits of the convictions, on all counts for which imprisonment has been imposed, that an appellate ruling for the defendants on any of those substantial questions would be likely to require a reversal of the conviction or a new trial. 18 U.S.C. § 3143(b)(2); United States v. Valera-Elizondo, 761 F.2d 1020 (5th Cir.1985).
The primary substantial question of law raised for appeal is whether the government obtained the defendants’ convictions through use of perjury and other false evidence, known to be such by the government’s representatives, that the government knowingly allowed to go uncorrected during the jury trial in which it appeared.
A principal element of the prosecution theory was that the defendants, under the leadership of Michael O’Keefe, Sr., caused a domestic insurer, Physicians National Risk Retention Group (PNRRG), which they managed, to enter a sham reinsurance contract with a foreign insurer, Builders and Contractors Insurance (BCI), managed by Charles Donaldson, who later became a key prosecution witness. One crucial issue in the case was whether the defendants had foreknowledge that Donaldson had no authority to enter the contract for BCI from its owners or directors. An FBI agent’s report of her interview of Donaldson, after he agreed to cooperate in-the investigation of the defendants, related that Donaldson said that Michael O’Keefe, Sr., had “suggested [to Donaldson] that BCI’s shareholders *291meeting minutes be altered [by Donaldson] to make it appear that Donaldson had authority to enter into the PNRRG/BCI contract.” Donaldson FBI 302 Report, quoted in O’Keefe I, 128 F.3d at 888. The government’s attorneys placed this same statement in the factual basis for Donaldson’s guilty plea in another district court to one count of mail fraud in exchange for his testimony against the defendants. Id. In that guilty plea proceeding the FBI agent testified that the factual basis accurately reflected what Donaldson had said in his interview and Donaldson testified that the factual basis accurately set forth what transpired between him and Mr. O’Keefe. Trial Transcript (Cross Examination of Charles Donaldson), March 12,1996, at 110.
During Donaldson’s direct testimony in the' present case, the prosecuting attorneys did not ask Donaldson any questions about his alteration of the BCI shareholders’ meeting minutes. Immediately prior to Donaldson’s direct examination by the government’s attorneys, the prosecution handed a copy of the FBI 302 report to the defense. During cross-examination, Donaldson at first denied that he had ever said that O’Keefe had suggested that Donaldson should alter the BCI minutes. Later during the cross-examination, however, Donaldson testified that he had told the FBI agent that O’Keefe suggested that he alter the minutes and that it was a false statement. Still later on cross, Donaldson testified that during his guilty plea proceeding in Baton Rouge, the FBI agent had taken the stand and, in giving the factual basis for his plea, repeated the false statement that Michael O’Keefe, Sr., had acted with Donaldson to alter the BCI minutes. Donaldson also testified in the present case that the FBI agent’s testimony at his guilty plea proceeding was correct. Trial Transcript (Cross Examination of Charles Donaldson) at 110.
A conviction obtained through use of false evidence, known to be such by representatives of the government, must fall under the Fourteenth Amendment. Napue v. Illinois, 360 U.S. 264, 269, 79 S.Ct. 1173, 3 L.Ed.2d 1217 (1959) (citing Mooney v. Holohan, 294 U.S. 103, 55 S.Ct. 340, 79 L.Ed. 791 (1935)). See Miller v. Pate, 386 U.S. 1, 87 S.Ct. 785, 17 L.Ed.2d 690 (1967)(false “blood” on shorts). “ ‘[T]he same result obtains when the State, although not soliciting false evidence, allows it to go uncorrected when it appears.’ ” Giglio v. United States, 405 U.S. 150, 154, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972) (quoting Napue, 360 U.S. at 269, 79 S.Ct. 1173); United States v. Agurs, 427 U.S. 97, 103, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976). “ ‘It is of no consequence that the falsehood bore on the witness’ credibility rather than directly upon defendant’s guilt. A lie is a lie, no matter what its subject, and, if it is in anyway relevant to the case, the district attorney has the responsibility and the duty to correct what he knows to be false and elicit the truth. * * * That the district attorney’s silence was not the result of guile or a desire to prejudice matters little, for its impact was the same, preventing, as it did, a trial that could in any real sense be termed fair.’ ” Napue, 360 U.S. at 269-70, 79 S.Ct. 1173 (quoting People v. Savvides, 1 N.Y.2d 554, 154 N.Y.S.2d 885, 136 N.E.2d 853, 854-55 (1956)).
A new trial is required if the false testimony could in any reasonable likelihood have affected the judgment' of the jury. United States v. Bagley, 473 U.S. 667, 678-79, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985); Giglio v. United States, 405 U.S. 150, 154, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972) (quoting Napue, 360 U.S. at 271, 79 S.Ct. 1173); Kirkpatrick v. Whitley, 992 F.2d 491 (5th Cir.1993). The fact that the jury was apprised of other grounds for believing that the government witness may have had an interest in testifying against the defendant does not turn what was otherwise a tainted trial into a fair one. Napue, 360 U.S. at 270, 79 S.Ct. 1173.
The government’s attorneys, in a sidebar conference during the trial, at oral argument on the motion for a new trial, and in affidavits filed with the motion for reconsideration, took the position that Donaldson had never told the FBI that O’Keefe urged him to alter the BCI minutes, that the FBI agent who interviewed Donaldson mistakenly thought Donaldson had accused O’Keefe of complicity in the alteration of the BCI minutes, that the FBI agent attributed that incorrect *292statement to Donaldson in the written FBI 302 report of the interview, that the government attorneys in New Orleans became aware of the false statement in the 302 prior to Donaldson’s arraignment, and that the government attorneys in New Orleans tried unsuccessfully to correct the false statement before a government attorney in Baton Rouge placed it in the factual basis for Donaldson’s guilty plea in the federal court for Middle District of Louisiana. See Trial Court’s Memorandum and Order (Aug. 15, 1996), at 64-70. Thus, the government was aware before trial that through its own fault Donaldson’s 302 FBI statement and guilty plea factual basis contained a false statement accusing O’Keefe of actively participating in the falsification of the BCI minutes in order to facilitate the BCI/PNRRG reinsurance contract; moreover, the government knowingly did not correct the inaccuracies in the FBI 302 Report prior to making the report available to the defense before Donaldson’s testimony. The government was also aware before trial that Donaldson’s guilty plea factual basis was of record and accessible to the defendants. Thus, there is a substantial argument that the government knew before trial that there was a reasonable likelihood that the defense would ask Donaldson about the false statement before the jury but that the defense would not know that the false statement originated with the FBI agent’s mistake, and not with what Donaldson actually told the FBI agent. Nevertheless, the government did not inform the court or the defense prior to trial or Donaldson’s testimony of the true nature of the false statement or its source. The government did not ask Donaldson any questions about the false statement or attempt to correct the 302 or the factual basis during Donaldson’s direct testimony. The government did not call the FBI agent or any of its attorneys to explain what was false and elicit the truth. The government was aware that Donaldson perjured himself several times during his cross examination by testifying that he had made the statement to the FBI agent, although he later said he was lying at the time, and although he also testified several times that he did not make the false statement to the FBI agent. Nevertheless, the government knowingly did not' ever in open court before the jury correct or explain Donaldson’s per-jurious testimony or the false statement in the 302 or the factual basis.
Applying the constitutional principles set forth by the Supreme Court in the cases cited above, it is clear that defendants’ appeal raises a substantial question whether their convictions and sentences must fall under the Fourteenth Amendment because (1) they were obtained through the use of perjured testimony and false evidence, known to be such by representatives of the government, Napue v. Illinois, 360 U.S. at 269, 79 S.Ct. 1173; Miller v. Pate, 386 U.S. 1, 7, 87 S.Ct. 785, 17 L.Ed.2d 690 (1967); Faulder v. Johnson, 81 F.3d 515, 519 (5th Cir.), cert. denied, 519 U.S. 995, 117 S.Ct. 487, 136 L.Ed.2d 380 (1996); Pyles v. Johnson, 136 F.3d 986, 996 (5th Cir.1998); (2) although the government may not have directly solicited the perjury or false evidence, it was responsible for the creation of the false statement, knew or should have known that the jury would be exposed to the false statement wrongfully indicating that O’Keefe participated in altering the BCI minutes, and knowingly allowed it to go uncorrected when it appeared before the jury, Giglio v. United States, 405 U.S. at 154, 92 S.Ct. 763; Napue, 360 U.S. at 269, 79 S.Ct. 1173; United States v. Agurs, 427 U.S. 97, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976); Faulder v. Johnson, 81 F.3d at 519; Pyles v. Johnson, 136 F.3d at 996; and (3) the perjury and false evidence could in any reasonable likelihood have affected the judgment of the jury. United States v. Bagley, 473 U.S. at 678-79, 105 S.Ct. 3375; Giglio, 405 U.S. at 154, 92 S.Ct. 763; Napue, 360 U.S. at 271, 79 S.Ct. 1173; Kirkpatrick v. Whitley, 992 F.2d 491, 497 (5th Cir.1993). See 2 Wayne R. LaFave & Jerold H. Israel, Criminal Procedure § 19.5, at 534 (1984) (“This obligation [of the prosecutor to disclose] requires that it not suborn perjury, not use evidence known to be false, and not allow known false testimony of its witnesses to stand uncorrected.”).
Consequently, there is a substantial question as to whether O’Keefe I clearly erred by applying incorrect principles of law to determine whether the government violated its *293duty to correct perjury and false evidence and, if so, whether the government’s violations were material. In O’Keefe I, the court stated that:
[1] Along with other circuits, we have limited material lies to those that occur as a part of the prosecution’s case. The prosecution has a duty only to refrain from knowingly presenting perjured testimony and from knowingly failing to disclose that testimony used to convict a defendant was false. Thus, when the defense elicits the alleged perjury on cross-examination, no material falsehood has occurred because the government has not itself knowingly presented false testimony. [Id. at 894 (citations and internal quotations omitted)].
[2] [W]e do not find that there is a reasonable probability that the jury would have reached a different outcome even had it been fully aware of all of the alleged inconsistencies and falsehoods in Donaldson’s testimony. As a result, the falsehoods were not material and no Napue deprivation of due process occurred. [Id. at 898].
It is arguable that the first O’Keefe I statement is clearly erroneous because it is contrary to the decisions of the Supreme Court and to previous panel opinions of this Circuit following the Supreme Court cases. In Napue and Giglio, both of which involved perjury by a -prosecution witness during his cross-examination by a defense attorney, see Napue, 360 U.S. at 267-68 & n. 2, 79 S.Ct. 1173; Giglio, 405 U.S. at 153, 92 S.Ct. 763, the Court held that a conviction obtained through use of false evidence, known to be such by representatives of the prosecuting government, must fall ás a violation of due process; and that “ ‘[t]he same result obtains when the State, although not soliciting false evidence, allows it to go uncorrected when it appears.’”. Giglio, 405 U.S. at 154, 92 S.Ct. 763 (quoting Napue, 360 U.S. at 269, 79 S.Ct. 1173). Prior to O’Keefe I this Circuit has consistently followed Napue and Giglio. See Pyles v. Johnson, 136 F.3d 986, 996 (5th Cir.1998)(“A state denies a criminal defendant due process when it knowingly uses perjured testimony at trial or allows untrue testimony to go uncorreeted.”)(quoting Faulder, 81 F.3d at 519 (citing Napue)). See Cordova v. Collins, 953 F.2d 167, 171 (5th Cir.1992).
The second O’Keefe I statement is also clearly erroneous. The Supreme Court has held that when a Napue violation occurs “[a] new trial is required if ‘the false testimony could ... in any reasonable likelihood have affected the judgment of the jury’.” Giglio, supra, at 154, quoting Napue, supra, at 271. Prior to O’Keefe I this Circuit’s panels adhered to the same materiality standard. Kirkpatrick v. Whitley, 992 F.2d at 497 (“[I]f the prosecutor has knowingly used perjured testimony or false evidence, the standard is considerably less onerous [than for Brady violations]: the conviction ‘must be set aside if there is any reasonable likelihood that the false testimony could have affected the jury’s verdict____’” [citing Bagley, 473 U.S. at 679 n. 9, 105 S.Ct. 3375 (citing Napue) ]); Moody v. Johnson, 139 F.3d 477, 484 (5th Cir.1998)(“It is well settled that the State is not permitted to present false evidence or allow the presentation of false evidence to go uncorrected, [citing Giglio, Napue and Mooney v. Holohan ]. However, if false evidence is presented by the prosecution at trial, a new trial is warranted only if the false testimony could have, in any reasonable likelihood, affected the jury’s determination.”).
One panel of this court may not overrule the decision of a prior panel of this court (absent an intervening decision to the contrary by the Supreme Court or the en banc court, of which there are none). Hogue v. Johnson, 131 F.3d 466, 491 (5th Cir.1997); Barber v. Johnson, 145 F.3d 234 (5th Cir.1998).
Consequently, the defendants’ appeal raises a substantial question as to whether an application of the correct controlling principles of law set forth in Napue, Giglio, Pyles, Faulder, Moody and Kirkpatrick by the court on direct appeal should require a reversal of the convictions and a new trial because the government, although not directly soliciting false evidence in court, allowed it to go uncorrected when it appeared in court. It is also cogently arguable that the record on direct appeal will show that the government’s violation of due process and the resulting *294harm to the substantial rights of the defendants were far more grievous than that depicted by the O’Keefe I opinion because (i) the government’s bad faith, fault or neglect created the false statements in the FBI 302 report and Donaldson’s guilty plea factual basis; (ii) the government took no steps to disclose the falsehoods to the court or the defendants, although it became aware of them before trial; (iii) the government was aware of the risk that the false evidence would be referred to by Donaldson in his testimony at trial before the jury; (iv) when this risk became a reality the government took no steps before the jury to fulfill its duty to correct what it knew to be false and elicit the truth for the trier of the facts; (v) had the government complied with its duty by correcting the false statements prior to trial, the jury would not have received any of the very incriminating false evidence that O’Keefe had personally participated in altering the BCI minutes (including evidence that Donaldson’s accusation against O’Keefe had been placed in an official FBI 302 report and that an FBI agent had testified to receiving such a statement from Donaldson in federal court); (vi) had the government complied with its duty, after the false statements were repeatedly mentioned during trial, to apprise the jury of the true facts, (that the statement was false, that Donaldson had never said O’Keefe participated in the fabrication of the BCI minutes, that when Donaldson testified that he had made that statement to the FBI he committed perjury, that the FBI agent was responsible for the creation of the false statement, and that the government knowingly had allowed the trial to go forward without disclosing the false statements to the court or the defendants), the jury might well have concluded that the government had either fabricated evidence or had been highly negligent in its investigation or preparation of the FBI report and the factual basis to which the FBI agent testified to in another federal court. There is a substantial question whether the truth about the government’s creation and perpetuation of the false statement likely would have affected the jury by shaking its confidence in the government’s whole investigation and prosecution. Also, the jury might well have concluded that Donaldson was willing to say anything to help the government’s case and to damage the defense, and to commit perjury to cover up the government’s mistakes. Accordingly, the jury might well have concluded that Donaldson’s testimony was thoroughly lacking in credibility or reliability; Thus, there is a substantial question whether there is a reasonable likelihood that the false evidence and testimony affected the judgment of the jury so that a new trial is required.
IV.
The defendants make an additional argument which also appears to raise a substantial question of law or fact. The defendants contend that they have newly discovered evidence that prior to. Donaldson’s guilty plea, the government allowed $45,000 which did not belong to Donaldson to be deposited to his credit in his lawyer’s trust account and withdrawn after Donaldson’s guilty plea and used by Donaldson as if -it were his own money to pay part of his restitution obligation. The defendants argue that by these transfers, the government gave Donaldson an additional substantial benefit which the government knowingly did not disclose in its plea agreement with Donaldson. -In the plea agreement, the government and Donaldson represented that the only benefits he would receive for his plea, cooperation and testimony against other persons were a 3-level decrease in his criminal offense level, and a possible recommendation of a downward departure in sentencing. On direct examination at defendants’ trial Donaldson identified his written plea agreement and testified that it was the agreement he had signed with the government. The government elicited from Donaldson the specific benefits he had been granted or promised by the government. The prosecuting attorney did not ask Donaldson, and he did not volunteer any information, about the $45,000 that the government allowed him to use to pay part of his restitution obligation. The defendants contend that the government’s introduction of Donaldson’s testimony about the plea agreement and the agreement itself, which was introduced as an exhibit in the defendants’ trial, contained false statements under oath by Donaldson *295and government representatives that misled the jury as to the quality and credibility of Donaldson’s testimony, and could have in reasonable likelihood affected the judgment of the jury. The defendants contend that the government therefore violated the Fourteenth Amendment under both Napue and Brady by presenting knowingly uncorrected perjury and false evidence at trial and failing to disclose impeachment evidence to defendants.
In this regard, a further substantial question of law or fact is raised as to whether O’Keefe I’s clearly erroneous intervention into the merits of the ease prevented the successor trial judge from allowing the defendants to present substantive evidence in support of their motion for a new trial.
My pretermitting discussion of other questions raised by defendants’ appeal does not indicate any opinion as to whether any of these is a substantial question, which if decided favorably to the defendants, would be likely to result in reversal.
Conclusion
Accordingly, I respectfully dissent from the majority’s granting of the government’s motion to stay the district court’s order releasing the defendants on bail pending the government’s appeal from that order. I believe the defendants’ appeal will raise substantial questions of law and that they should therefore be allowed to remain on bail pending their own appeal. However,. I do not intend to intimate how the substantial questions raised by defendants’ appeal and the government’s argument based on the law of the case doctrine should be decided. It would be inappropriate to express an opinion on the merits of these questions prior to full briefing and presentation to this court on direct appeal. United States v. Clark, 917 F.2d 177, 181 (5th Cir.1990). The present panel’s function is merely to decide whether to stay the district court’s decision to grant bail pending the government’s appeal from the district court’s order releasing the defendants on bail pending their own appeal.
. The panel concluded that “harmless error" existed because: (i) little risk of injustice would result from not vacating the denial of the motion for reconsideration and remanding the case to the successor judge, who had been assigned the case, for a decision on the government’s reconsideration motion; (ii) a decision on the merits of the trial judge’s granting of the defendants’ motion for new trial would serve justice in other cases because it would clarify an unclear area of the law and admonish district judges as to the importance of taking no discretionary actions after recusal; and (iii) there is little risk of undermining the public’s confidence in the judicial process. O’Keefe I, 128 F.3d at 892-93.
. The O'Keefe I panel also found that two of the additional findings were inherently flawed: (i) the government's delay in disclosing the FBI 302 reports of investigative interviews of the two key prosecution witnesses, Donaldson and Moore, containing exculpatory evidence, did not violate Brady by impairing defendants' ability to cross-examine those witnesses, based on the panel’s review of the record and the absence of any affirmative finding (other than the conclusion) by the district judge to that effect, O’Keefe I, 128 F.3d at 898-99; and (ii) even if the prosecution attempted to mislead the defense by redrafting of the indictment, the defense had too much knowledge of the underlying facts to be misled. Id. at 895-96, 899.
. Doddy v. Oxy USA, Inc., 101 F.3d 448 (5th Cir.1996), although problematic, is not inconsistent with the general rule. The trial judge, who owned Exxon stock, recused herself when informed that a corporate party had become affiliated with that company. Later the same day, she vacated the order of recusal and took evidence on the relationship of the two corporations. The evidence indicated that the two corporations had formed a joint venture but that neither Exxon nor the third entity venture could be affected by the litigation. The trial judge referred the question of recusal to the chief judge of the district, who determined that there was no basis for recusal. Under these circumstances, although technically the trial judge performed a discretionary act in vacating her own recusal, the chief judge could and would have vacated the recusal in any event, and her act had no effect upon the parties or the litigation. The trial judge did not deliberately violate her own recusal order by fully stepping back into role of trial judge and talcing further discretionary action in disregard or violation of her initial recusal. The trial judge’s infraction of § 455, if any, was tire kind of "harmless error committed by busy judges who inadvertently overlook a disqualifying circumstance” that the Supreme Court indicated "there is surely room for.” Liljeberg v. Health Services Acquisition Corp., 486 U.S. 847, 862, 108 S.Ct. 2194, 100 L.Ed.2d 855 (1988).
Moreover, under 28 U.S.C. § 455(f), added in 1988, after Liljeberg, the trial judge in situations like Doddy v. Oxy USA, Inc., supra, may be able to avoid disqualifying herself altogether. Section 455(f) provides that if a judge, after substantial judicial time has been devoted to a matter, discovers or is apprised that she has a financial interest in a party (other than an interest that could be substantially affected by tire outcome), disqualification is not required if the judge divests herself of that financial interest.
In re Continental Airlines Corp., 901 F.2d 1259, 1263 (5th Cir.1990) is distinguishable because it did not involve a judge's deliberate violation of his own disqualification order.
. See 11 Charles Alan Wright, Arthur R. Miller, & Mary Kay Kane, Federal Practice and Procedure § 2857, at 254-57 (2d ed.1995). "A number of cases say that discretion ordinarily should incline toward granting rather than denying relief, especially if no intervening rights have attached in reliance upon the judgment and no actual injustice will ensue. The policy of the law to favor a hearing of a litigant’s claim on the merits must be balanced against the desire to achieve finality in litigation.” Id. at 255-57. (footnotes omitted).
. In general terms, harmless error analysis usually calls upon a reviewing court, either trial or appellate, to set aside a verdict or to disturb a judgment when an error affects the substantial rights of the unsuccessful party, and to disregard any error which does not, leaving the judgment undisturbed. See Fed.R.Civ.P. 61 and Fed. R.Crim.P. 52(a). It is doubtful that any verbal formulation can avoid the subjectivity that necessarily inheres in determining whether an error has affected the substantial right of a party. II Charles Alan Wright, Arthur R. Miller, Mary Kay Kane, Federal Practice and Procedure § 2883, at 445-46 (2d ed.1995). Two of the best short attempts have been made by Judge Traynor and Justice Rutledge. See Roger Traynor, The Riddle of Harmless Error 35 (1970)("[U]nless the appellate court believes it highly probable that the error did not affect the judgment, it should reverse.”); and Kotteakos v. United States, 328 U.S. 750, 760, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946) (Rutledge, J.) (“Do not be technical, where technicality does not really hurt the party whose rights in the trial, and in its outcome the technicality affects.”), cited in 11 Charles Alan Wright, Arthur R. Miller, Mary Kay Kane, Federal Practice and Procedure § 2883, at 445-47. Thus, a harmless error determination is concerned mainly with whether an error has affected the outcome of a case to the substantial disadvantage of the losing party.
On the other hand, exercising discretion in deciding whether to relieve a party from a final civil judgment under Rule 60(b)(6) because of a judge’s § 455(a) violation does not involve a determination of whether the violation affected the outcome of the case. Instead, it involves the consideration of a more complex constellation of factors, including the impact that denying or granting relief will have upon the values of justice for both parties, the deterrence of judicial misconduct, the appearance of justice and the integrity of the courts. Consequently, the considerations mentioned in Liljeberg, i.e., risks of injustice to die parties, injustice in odier cases, and the undermining of confidence in the courts, are pertinent and useful to the exercise of judicial discretion under Rule 60(b)(6) and § 455, but they are not particularly relevant or helpful to a harmless error determination.
. The harmless error rule calls upon a reviewing court to (1) disregard a harmless error, viz., one that does not really hurt the complaining party or affect the outcome of the judgment or order complained of, and (2) leave undisturbed, i.e., to affirm, an order or judgment affected only by harmless error. See 11 Charles Alan Wright, Arthur R. Miller, & Mary Kay Kane, Federal Practice and Procedure % 2883; 12 Moore’s Federal Practice § 61.02 (3d ed.1998). O'Keefe I used the equitable factors to take the first step in applying a harmless error rule by disregarding die trial judge’s error of law in violating his own disqualification order, but O’Keefe I did not follow through with the second step required by a harmless error rule, i.e., affirming the. trial judge’s ruling on the motion for reconsideration and remanding for further proceedings. Instead, O’Keefe I left the second part of its so-called harmless error operation undone, skipped over the usual step of affirming an order or judgment free of harmful error, to a review of die merits of the trial judge’s ruling on the motion for a new *288trial. Based on that merits review O’Keefe I vacated the new trial order and, sub silentio, vacated the trial judge’s ruling on the motion to reconsider that it had earlier found to be affirma-ble as harmless error, and denied the defendant's motion for new trial on the merits and with prejudice. Thus, the convolutions and non se-quiturs involved in O'Keefe /’s application of what it mischaracterized as a harmless error rule underscore that there is a substantial question as to whether that panel’s decision was clearly erroneous.