After Chief Judge Cerezo of the United States District Court for the District of Puer-to Rico set a firm trial date in a case presently pending before her, United States v. Lorenzo Munoz-Franco, 14 F.Supp.2d 167 (D.Puerto Rico 1998) the government moved at the eleventh hour to disqualify the judge from further involvement. The judge denied the motion following a three-day evidentiary hearing. The government then sought a writ of mandamus from this court directing Judge Cerezo to recuse herself. We provisionally stayed the impending trial, set an expedited briefing schedule, and entertained oral argument. We now conclude that the government failed to prove what it had alleged visa-vis the judge, and therefore deny the petition.
At the outset, it is important to note the narrowness of the government’s position: it does not contend that the judge has any actual bias or prejudice in this case and it does not seek her recusal under 28 U.S.C. § 144 (1994). It likewise eschews the mandatory bases for disqualification limned in 28 U.S.C. § 455(b) (1994). Instead, the government premises its mandamus petition (and the underlying recusal motion) exclusively on 28 U.S.C. § 455(a) (1994), which provides:
Any justice, judge, or magistrate of the United States shall disqualify himself in any proceeding in which his impartiality might reasonably be questioned.
*28In cases involving section 455(a), the recu-sal determination inevitably turns on the facts. See Liljeberg v. Health Servs. Acquisition Corp., 486 U.S. 847, 865, 108 S.Ct. 2194, 100 L.Ed.2d 855 (1988). Consequently, we describe the pertinent events in some detail. We then discuss the applicable law and, finally, undertake an analysis of the recusal question.
I. BACKGROUND
United States v. Munoz-Franco stems from the May 1990 failure of Caguas Central Federal Savings Bank (Caguas), reputed to be the largest bank failure in the history of Puerto Rico. The government tells us, without demurrer by the respondents, that Cag-uas’s collapse resulted in aggregate losses exceeding $120,000,000.
The defendants in Munoz-Franco include two former Caguas officials, namely, Lorenzo Muñoz Franco (Muñoz), Caguas’s chief executive officer, and Francisco Sánchez Arán (Sánchez), Caguas’s chief lending officer.1 The indictment charges Muñoz and Sánchez with misapplying bank funds, making false entries in banking records, and participating in a conspiracy to perpetrate these offenses and to commit bank fraud. See 18 U.S.C. §§ 371, 657, 1006, & 1344 (1994). In its narrative portions, the indictment describes a “loan-kiting” scheme that purportedly involved the misapplication of real estate loan proceeds to shore up other (failing) commercial loans, thereby creating the illusion that the latter loans were performing well. The government alleges that one object of the scheme — which supposedly persisted for almost the entire decade between 1980 and 1990 — was to stave off regulatory intervention and keep Muñoz and Sánchez in power.
The transaction upon which the government bases its recusal initiative took wing in 1986 when the judge’s husband, Benny Frankie Cerezo, sought to borrow funds from Caguas. Mr. Cerezo approached Arturo So-mohano, Caguas’s senior vice-president for commercial lending, and explained that he wished to obtain a loan so that he could develop a twenty-eight acre farm and subdivide it into house lots. The record is teneb-rous as to whether Mr. Cerezo furnished appraisal reports in support of the loan application, but we do know that he at least provided Caguas with the cover letters from two appraisal reports prepared in 1984. Both letters subscribed that the acreage had a value of $200,000 or more.
Despite the fact that Mr. Cerezo’s checking account was overdrawn,2 Somohano approved the application and the Cerezos obtained a $150,000 loan from Caguas in the autumn of 1986 at two points over prime, secured by a first mortgage on the farm. The loan contract and related documents were signed by Mr. Cerezo (individually and on behalf of his wife, via power of attorney). The promissory note called for eleven monthly interest payments and repayment of the loan principal on the first anniversary. Between November 1986 and November 1987 (when the loan matured), the Cerezos made at most three interest payments.
As the note neared maturity, Mr. Cerezo requested a loan of $557,000 as additional financing for his shoe business. He initially made this request in a Spanish-language letter to Muñoz, dated October 26, 1987. The salutation of the letter read “Estimado amigo Lorenzo” (“Esteemed friend Lorenzo”), but the body of the letter employed formal verb forms (conjugated for use with “usted” rather than with the more familiar “tu”). Muñoz referred the letter to Somohano and, three days later, Mr. Cerezo wrote directly to So-mohano, making essentially the same request and indicating that the earlier letter to Mu-ñoz had been sent in error. Caguas never approved the $557,000 loan.
Upon maturity, the Cerezos failed to repay the farm loan. During the initial post-default *29period, which extended from November 1987 (when the note matured) until May 1990 (when control of the note passed into the hands of third parties, see infra), the record does not reflect that either Muñoz or Sán-chez had anything to do with Caguas’s collection efforts. We review what transpired.
Over the first eight months of the post-default period, Caguas sent Mr. Cerezo three collection letters, each of which demanded immediate payment. Somohano sent a copy of the second letter to Judge Cerezo at the Cerezos’ home address because he was concerned that, as a cosigner by power of attorney, she might not have been aware that the loan even existed. The letters did not accomplish their intended purpose. Mr. Cere-zo informed Caguas that he did not have liquid funds sufficient to repay the debt. He proposed several alternatives, such as working out a plan to sell the farm or reviving his application for a loan to finance the expansion of his shoe business (combining the existing loan with the new loan). Because So-mohano terminated his employment with Caguas shortly after sending the second collection letter, these suggestions were considered by Pedro Suau, Caguas’s assistant vice-president for commercial lending. Suau countered with a proposal to rewrite the farm loan for $185,000 in order to cover the accrued interest and create a reserve for interest payments over the following six months. Mr. Cerezo displayed no enthusiasm for this proposal and faxed a letter to Suau in October 1988, with a copy to Muñoz, urging approval of his $557,000 loan request. Neither the $557,000 “shoe business expansion” loan nor the $185,000 refinancing of the farm loan ever materialized.
By February 1989, the Cerezos owed a total of $185,873.57 in principal plus accrued interest on the delinquent farm loan. At that juncture, Suau recommended transferring the matter to the bank’s special loans department (which handled workouts). The transfer did not occur at that time, however, as the document bearing Suau’s recommendation also carried an undated, unsigned, handwritten notation that stated: “This case was not authorized to go or to pass into the special department.” The next month, Mr. Cerezo sent $5,000 to Caguas in partial payment of accrued interest on the delinquent farm loan. He made no further payments, and Caguas eventually transferred the loan to the workout section. It was then referred to the bank’s outside counsel, who sent a dunning letter to Mr. Cerezo (with a copy to Judge Cerezo) on November 21,1989.
Mr. Cerezo asked the lawyers for a ninety-day extension, stating that, if he could not sell the farm for an amount sufficient to liquidate the debt within that interval, he would deed the property to Caguas as payment in kind. The law firm approved the extension request, but nothing happened. The attorneys fired off another collection letter. In response, Mr. Cerezo proposed surrendering the farm as payment in kind. Caguas indicated that it would consider the proposal on condition that the Cerezos secure a satisfactory appraisal of the property. Mr. Cerezo acquiesced to this condition and, in May 1990, he retained an appraiser and advised Caguas of the appraiser’s identity.
On May 25, 1990, Caguas failed. The Resolution Trust Corporation (RTC) moved into the picture, first as conservator and later as receiver. On August 31, the RTC and Banco Santander Puerto Rico (Santan-der), an unrelated third party, signed a purchase and assumption agreement pursuant to which Santander bought various assets of Caguas, including the farm loan. On February 22, 1991, Santander’s attorneys advised Mr. Cerezo that Santander had acquired the note and mortgage, and was willing to accept the farm as payment in kind. Mr. Cerezo did not respond to this communique.
On April 3, 1991, Santander brought a foreclosure complaint against the Cerezos, alleging an aggregate debt of $220,175.34 through March 28,1991. Process was served on both Mr. Cerezo and Judge Cerezo during June of that year. The Cerezos did not contest the allegations and a judgment of foreclosure entered on October 28, 1991. At the ensuing public auction, held on May 6, 1992, the farm was sold to CREFISA, San-tander’s real estate subsidiary. CREFISA held the property for a time and, in 1994, sold it to ah unrelated third party, in an arm’s-length transaction, for $92,000. San-*30tander never sought to obtain a deficiency-judgment against the Cerezos, apparently in keeping with its custom of not pursuing deficiencies on defaulted real estate loans. Neither Muñoz nor Sánchez had any relationship with Santander, and neither of them participated in any way in the decision not to pursue the deficiency.
II. APPLICABLE LEGAL STANDARDS
Section 455(a) stands at a crossroads where competing policy considerations frequently intersect. On the one hand, “courts must not only be, but must seem to be, free of bias or prejudice.” In re United States, 666 F.2d 690, 694 (1st Cir.1981). On the other hand, recusal on demand would put too large a club in the hands of litigants and lawyers, enabling them to veto the assignment of judges for no good reason. Thus, compulsory recusal must require more than subjective fears, unsupported accusations, or unfounded surmise. See id.
Section 455(a) attempts to reconcile these competing policies. The statute requires a judge to step down only if the charge against her is supported by a factual foundation and “the facts provide what an objective, knowledgeable member of the public would find to be a reasonable basis for doubting the judge’s impartiality.” Id. at 695 (emphasis in the original). While doubts ordinarily ought to be resolved in favor of recusal, see Nichols v. Alley, 71 F.3d 347, 352 (10th Cir.1995); United States v. Dandy, 998 F.2d 1344, 1349 (6th Cir.1993), the challenged judge enjoys a margin of discretion:
[T]he analysis of allegations, the balancing of policies, and the resulting decision whether to disqualify are in the first instance committed to the district judge. And, since in many cases reasonable deciders may disagree, the district judge is allowed a range of discretion. The appellate court, therefore, must ask itself not whether it would have decided as did the trial court, but whether that decision cannot be defended as a rational conclusion supported by [a] reasonable reading of the record.
In re United States, 666 F.2d at 695.
Inasmuch as this matter comes to us by way of a petition for a writ of mandamus, two additional sets of considerations are implicated. The first is surmountable. For both prudential and practical reasons, we must be slow to “foster piecemeal review and disturb the historic relationship between trial and appellate courts.” In re Cargill, Inc., 66 F.3d 1256, 1259 (1st Cir.1995), cert. denied, 517 U.S. 1156, 116 S.Ct. 1545, 134 L.Ed.2d 648 (1996). It is only in unusual situations that interlocutory review of a judge’s refusal to step aside is available through mandamus. See In re Martinez-Catala, 129 F.3d 213, 217 (1st Cir.1997).
However, that principle is not ironclad. When “the issue of partiality has been broadly publicized, and the claim of bias cannot be labelled as frivolous,” the propriety of judicial disqualification need not await end-of-case review. In re United States, 666 F.2d at 694. This is especially true in a criminal case in which the government seeks the judge’s recusal, for a defendant’s verdict will terminate the case, thereby rendering the usual remedy, end-of-case appeal, illusory. Cf. United States v. Patterson, 882 F.2d 595, 599-600 (1st Cir.1989) (holding that mandamus was an appropriate avenue for government to obtain review of district court determination that prior convictions did not qualify as predicate offenses since government could not appeal from an ensuing sentencing order). Here, the issue of recusal vel non has attracted widespread interest and, on its face, the government’s allegation cannot be termed frivolous. Thus, mandamus review appears proper.
The second set of considerations relates to the fact that mandamus has its own ingrained jurisprudence. An applicant for the writ “must show both that there is a clear entitlement to the relief requested, and that irreparable harm will likely occur if the writ is withheld.” In re Cargill, 66 F.3d at 1260. Taking these elements in reverse order, the government’s claim of irreparable harm has obvious weight. Unlike situations in which a claim of irreparable harm is undercut by the availability of full end-of-case review, see, e.g., id. at 1264 n. 10, the government here will have no effective means of *31correcting the judge’s alleged error&emdash;her failure to recuse herself&emdash;on appeal if the case is tried and the defendants prevail.
In our view, these same circumstances also affect the application of the “clear entitlement” requirement. In the run-of-the-mine recusal case, “mandamus is almost always withheld-we do not say always-unless the petitioner demonstrates that it is ‘clearly’ entitled to relief.” In re Martinez-Catala, 129 F.3d at 218. In other words, “mandamus requires a case not merely close to the line but clearly over it.” Id. at 221. We believe it is questionable whether this requirement should be applied stringently when the government seeks a judge’s disqualification in a criminal case. Because of the government’s inability to press an end-of-case appeal, see supra, we think it would be fairer in such circumstances to use the ordinary abuse-of-discretion standard rather than the more exacting standard usually applicable to petitions for mandamus. We follow this course.
III. THE MERITS
The question before us is not whether we would have denied the government’s recusal motion as Judge Cerezo did, but, rather, whether her denial of it constituted an abuse of discretion. The answer to this question depends on whether a reasonable person, fully informed of all the facts, would doubt Judge Cerezo’s impartiality. The standard, of course, is objective, not subjective.
A
Typically, cases implicating section 455(a) are fact-specific, and thus sui generis. Comparison, therefore, is an inexact construct. Nonetheless, a rough continuum of sorts emerges from a study of the case law. At one end are situations in which the hypothesis of partiality is so compelling that the judge has no real choice but to recuse herself. See, e.g., Fredonia Broad. Corp. v. RCA Corp., 569 F.2d 251 (5th Cir.1978) (involving a recusal motion based upon one party’s representation by the judge’s former law clerk, who had served in that capacity during a prior trial of the same action). At the other end are situations in which the hypothesis of partiality is so tenuous that the judge has no real choice but to sit. See Blizard v. Frechette, 601 F.2d 1217 (1st Cir.1979) (involving a recusal motion based upon nothing more than a judge’s criticism of a party and her case in an opinion). Between these two polar extremes lies a zone in which the district judge’s discretion holds sway. See In re United States, 666 F.2d at 695. If a case falls within this gray area, a court of appeals ought not to interfere.
These categories operate on two levels: what is alleged and what is proven. See Blizard, 601 F.2d at 1221 (“A trial judge must hear cases unless some reasonable factual basis to doubt the impartiality or fairness of the tribunal is shown by some kind of probative evidence.”); cf. Liljeberg, 486 U.S. at 865, 108 S.Ct. 2194 (admonishing that “it is -critically important in a case of this kind to identify the facts that might reasonably cause an objective observer to question [the judge’s] impartiality”). Consequently, there are situations in which the hypothesis of partiality sounds compelling to a reasonable listener, but the supposed facts upon which the hypothesis depends are, in the end, not proven. Without facts to substantiate a hypothesis of partiality, a case may well slide from one pole of the continuum to the other, or to some point in between. The case at hand is a perfect illustration of this phenomenon&emdash;and it is for this reason that the mandamus petition fails.
B
As we have explained, section 455(a) requires that we ask whether a reasonable person, fully informed of all the relevant facts, would fairly question the trial judge’s impartiality. The facts proven below must dictate the answer to this question, not the unverified suspicions harbored by the government or the innuendo interspersed throughout both its recusal motion and its mandamus petition.
Stripped of the government’s rhetorical gloss, the relevant facts are as follows. The Cerezos had a delinquent loan at Caguas during a period in which Muñoz and Sánchez headed that bank and allegedly engaged in a *32fraudulent loan-kiting scheme. There is absolutely no proof that Muñoz or Sánchez had anything to do with approval of the Cerezos’ loan. By like token, there is no basis for suggesting that the Cerezos secured the loan by undue influence or other improper means. Their combined earning power and net worth augured creditworthiness; the bank obtained what appeared at the time to be adequate collateral; the interest rate (two points over prime) did not smack of favoritism; and So-mohano’s uncontradieted testimony at the re-cusal hearing verified that nothing in the loan documents indicated any irregularity. While Mr. Cerezo’s checking account was underwater at the time, the amounts of the overdrafts were relatively modest, and'the record does not contain any evidence that Caguas rejected other prospective borrowers because of modest checking account overdrafts. Similarly, though Caguas may have used bad business judgment in granting the farm loan, no adverse inference reasonably can be drawn ficom that fact; had Caguas dealt only with rock-solid borrowers and exercised prudence overall, the bank probably would not have failed.
The record is likewise bereft of any evidence that Caguas deviated from regular practice in its handling of the loan, or that Muñpz and Sánchez were involved in the loan’s administration after the initial default. There is not a shred of proof that Caguas exhibited less tolerance toward other borrowers who missed interest payments or that Caguas treated other “problem” loans more aggressively.3 The same is true as to Cag-uas’s actions following nonpayment of the principal. Although a total of almost four years elapsed between default (November 1987) and foreclosure (October 1991), the relevant period is slightly over two-and-one-half years. Caguas failed in May of 1990, and anything that transpired after that date had nothing to do with Muñoz and Sánchez. From that point forward, unrelated third parties — namely, the RTC and Santander— were in the driver’s seat.
Moreover, Caguas did not let the matter lie fallow during the post-default period. It promptly sent a series of letters in an effort to collect the debt. Following this battery of letters and a number of related telephone calls, the two sides exchanged proposals for resolution of the impasse.4 The other pan of the scale is altogether empty: the government adduced no evidence that Caguas employed different stratagems or greater diligence in respect to other defaulted commercial loans.
An unknown person did overrule Suau’s decision to transfer the delinquent loan to the workout section during this period, but that fact, standing alone and unexplained, proves nothing. The government neither called Suau as a witness nor procured an affidavit from him. It adduced no evidence tracing the undated, unsigned, handwritten notation that upstaged Suau to either Muñoz or Sánchez. And, finally, it did not show that *33Caguas customarily shifted loans of a similar age, size, and state to the workout section.
After Mr. Cerezo made a $5,000 payment on account, Caguas referred the loan to outside counsel for collection. The Cerezos then were granted an additional ninety-day moratorium. The record does not indicate that this sort of extension was unusual, or that it betokened any sort of special favoritism, or that either criminal defendant was involved in the extension decision. Mr. Cerezo next agreed to obtain a new appraisal before deeding the farm to Caguas in lieu of foreclosure. At that point, Caguas’s insolvency intervened.
Among the witnesses who testified, those best positioned to know the details of what had happened were Somohano and Angel Alicea Parés (Caguas’s outside counsel). After reviewing all the loan documents, Somo-hano testified without contradiction that he did not believe that Caguas afforded the Cerezos any preferential treatment. He also testified that the loan appeared to have followed Caguas’s normal collection procedures. In respect to the period after Caguas referred the loan to outside counsel (but before the conservatorship attached), Caguas’s collection attorney testified in a similar vein: no one at Caguas asked him to accord kid-glove treatment to these debtors, and he accorded none. He also vouchsafed that there was nothing out of the ordinary in the manner in which he proceeded in respect to this loan. This testimony, too, was unrefuted.
In its only real effort to fill the evidentiary void, the government notes that Mr. Cerezo addressed a request for additional funding to his “friend,” Muñoz. The letter itself is at best ambiguous as to the extent of any friendship. Moreover, despite the government’s claim of a sinister alliance, there is no explanation of what relationship, if any, existed between the two men. Such an undeveloped suggestion, in and of itself, does not mandate recusal. See In re Beard, 811 F.2d 818, 828 (4th Cir.1987) (“ ‘Mere general allegations of intimacy of the judge with opponents’ are insufficient to require recusal ....”) (quoting Morse v. Lewis, 54 F.2d 1027, 1031 (4th Cir.1932)); TV Communications Network, Inc. v. ESPN, Inc., 767 F.Supp. 1077, 1079 (D.Colo.1991) (“Mere allegations of a social relationship between a judge and a litigant in his court are not sufficient grounds for disqualification.”);. cf. In re United States, 666 F.2d at 696-97 (“If the receipt by a judge’s friend of a favor long ago from one who is a present litigant should disqualify the judge, judges could hope to preside without challenge solely in communities in which they are strangers.”).
To sum up, there is simply no basis for a founded conclusion that the Cerezos received any sort of preferential treatment or that Caguas treated the Cerezos differently than any other borrowers. The record is equally barren of any evidence that either of the criminal defendants dealt with Mr. Cerezo, had any involvement with the approval or administration of the farm loan (either before or after the Cerezos defaulted), or had anything whatever to do with Santander’s decision not to pursue the Cerezos for the deficiency that resulted from its post-foreclosure sale of the property. On this exiguous record, it is surpassingly difficult to see how the government can prevail.
C
In a last-ditch effort to overcome the paucity of its proof, the government maintains that the district judge’s actions during the recusal hearing provide grounds for disqualification. This is new matter, for the government never made this claim in the district court and probably has waived it. In all events, its arguments sound suspiciously like arguments for the proposition that the judge exhibited an actual bias against the government. This does not square with the government’s repeated assertions that its case for recusal in no way involves charges of actual bias, but hinges strictly and solely on a professed appearance of bias.
Passing this point and turning to what transpired at the hearing, we note first that the substance of this recusal motion — which the government hardly can be faulted for bringing — touched upon an extremely sensitive subject, involving the financial difficulties of the judge and her husband. Under such circumstances, we think that the judge would *34have been well-advised either to bow out of the case or to ask that the recusal motion be assigned to a different judge for hearing. Still, the law does not require a judge to step aside whenever a litigant raises a sensitive subject — if it did, litigants would have an easy means of disqualifying judges who were not to their taste — and this case is no exception to that rule. Indeed, recusal motions under section 455(a), which customarily are decided by the judge whom the movant seeks to disqualify, almost always involve the actions or relationships of the judge and almost always require the judge to appraise her own situation. See In re Martinez-Catala, 129 F.3d at 220.
We note, too, that the court’s inquiry into the timeliness of the government’s recusal motion — itself a proper subject for scrutiny, see In re Abijoe Realty Corp., 943 F.2d 121, 126 (1st Cir.1991) — necessarily involved an exploration into when and how the government sought to substantiate the information it had received about a delinquent Cerezo loan.5 The judge expressed understandable concern about several abortive attempts by the government to obtain the loan file sub rosa, by employing various artifices, including the issuance of a subpoena, later withdrawn, that did not comply with the notification requirements of the Right to Financial Privacy Act, 12 U.S.C. §§ 3401-3422 (1994). It is evident that the judge’s questioning went too far, and that its tone at times was overly confrontational. Yet, we find no basis for concluding that an irredeemable conflict existed between the judge’s judicial role and her concerns as a bank customer. As one member of the Court recently observed,
§ 455(a) is triggered by an attitude or state of mind so resistant to fair and dispassionate inquiry as to cause a party, the public, or a reviewing court to have reasonable grounds to question the neutral and objective character of a judge’s rulings or findings. I think all would agree that a high threshold is required to satisfy this standard. Thus, under § 455(a), a judge should be disqualified only if it appears that he or she harbors an aversion, hostility or disposition of a kind that a fair-minded person could not set aside when judging the dispute.
Liteky v. United States, 510 U.S. 540, 557-58, 114 S.Ct. 1147, 127 L.Ed.2d 474 (1994) (Kennedy, J., concurring in the judgment). We do not believe that the manner in which Judge Cerezo conducted the evidentiary hearing, though inappropriate in certain respects, revealed anything that would necessitate her recusal under these, criteria.
Our dissenting brother raises an argument that the government eschews. He charges that Judge Cerezo abused her discretion by hearing the recusal motion. See post at 37 & n. 10. This is a startling proposition, unsupported by any authority — and all the more startling because no party to the ease suggested below that a different judge should hear the motion. Although a trial judge faced with a section 455(a) recusal motion may, in her discretion, leave the motion to a different judge, see United States v. Heldt, 668 F.2d 1238, 1271 & n. 69 (D.C.Cir.1981) (per curiam); 13A Charles Alan Wright, Arthur R. Miller, & Edward H. Cooper, Federal Practice and Procedure § 3550 (2d ed.1984), no reported case or accepted principle of law compels her to do so — especially where, as here, all the litigants apparently are content to .have her hear the motion. We conclude, therefore, that Judge Cerezo acted within her proper role in conducting the hearing herself.6 See In re Martinez-Catala, 129 F.3d at 220; In re United States, 666 F.2d at 695.
D
This is a vexing case. In its recusal motion, the government painted a picture of *35coziness and preferential treatment which, if proven, plainly would have created an appearance of partiality and thus have demanded recusal. But those allegations were not proven. Then, in its mandamus petition, the government renewed its charges, but with variations designed to account for the wide disparity between what it had alleged and what it had succeeded in proving. Like the government’s original scenario, these rhetorical improvisations do not withstand scrutiny.
In the end, we are left with nothing more than these few facts: the Cerezos, during a small part of the relevant decade, borrowed money from Caguas at a standard rate and on conventional terms; they were unable to repay the loan; and the bank, acting with a ponderousness that frequently characterizes large bureaucratic institutions, took its time about instituting collection proceedings. Apart from the loan itself, the Cerezos received no special benefit from Caguas — and there is no proof that Muñoz or Sánchez either facilitated the loan or contributed to the delays that marked the collection process. These facts do not transport the case into the zone of obligatory recusal, but, rather, bring it within the “range of discretion” in which a decision either way can “be defended as a rational conclusion supported by [a] reasonable reading of the record.” In re United States, 666 F.2d at 695. Thus, although Judge Cerezo could have chosen to withdraw — indeed, we think that may have been the wiser course and that many judges would have taken it — and may yet choose to do so, she was not duty bound to disqualify herself from presiding over the criminal trial.
This conclusion is reinforced by the history of Caguas’s difficulties. The RTC brought a civil suit in May 1993 against Muñoz, Sán-chez, and several former directors of Caguas, alleging that these persons caused the extravagant losses that Caguas experienced. Judge Cerezo presided over this litigation from the start and continues to do so. A bench trial is scheduled to commence before her on May 11, 1999. The government, which is a party to that parallel litigation, has never sought her recusal, and apparently is satisfied that the judge’s participation does not give rise to any appearance of impropriety. The government cannot have it both ways. See United States v. Tierney, 760 F.2d 382, 388 (1st Cir.1985) (“Having one’s cake and eating it, too, is not in fashion in this circuit.”).
The other asseverations advanced by the government and the dissent require no further response. It suffices to say that we have given deliberate consideration to the claims raised in this mandamus petition, but find them largely unproven and therefore unpersuasive. Just as a judge must assiduously avoid participating “in any proceeding in which h[er] impartiality might reasonably be questioned,” 28 U.S.C. § 455(a), so, too, a judge must avoid yielding in the face of unfounded insinuations. A party cannot cast sinister aspersions, fail to provide a factual basis for those aspersions, and then claim that the judge must disqualify herself because the aspersions, ex proprio vigore, create a cloud on her impartiality. See FDIC v. Sweeney, 136 F.3d 216, 219 (1st Cir.1998) (per curiam); 13A Charles Alan Wright, Arthur R. Miller & Edward H. Cooper, Federal Practice and Procedure § 3542 (2d ed.1984). To hold otherwise would transform recusal motions into tactical weapons which prosecutors and private lawyers alike could trigger by manipulating the gossamer strands of speculation and surmise.
We need go no further. On this meager record, Judge Cerezo acted within her discretion in concluding that a reasonable person, fully informed of all the facts, would not fairly question her impartiality. Consequently, the government has failed to establish that the judge’s failure to remove herself from the pending criminal case warrants our intervention.7
The petition for writ of mandamus is denied and dismissed.
. The remaining defendants are Ariel and Enrique Gutiérrez, quondam commercial customers of Caguas, their employee, Wilfredo Umpierre-Hernández, and a real estate developer, Rafael Dominguez-Wolff. The government does not claim that the involvement of any of these individuals creates (or contributes to) the perceived need for Judge Cerezo to step aside. Thus, we concentrate solely on Muñoz and Sánchez.
. Mr. Cerezo's checking account remained overdrawn in amounts ranging from $2,329.51 to $2,748.28 during the ensuing year.
. The government might respond that it presented no such proof because it did not fully apprehend the loan's nuances until the contents of the entire loan file surfaced at the recusal hearing. There are two obvious rejoinders to such a disclaimer. First, because the government bore the burden of making the case for recusal, see Denardo v. Municipality of Anchorage, 974 F.2d 1200, 1201 (9th Cir.1992); Pope v. Federal Express Corp., 974 F.2d 982, 985 (8th Cir.1992), the lack of proof ineluctably counts against it. Second, if the government was surprised at the hearing— and we note, parenthetically, that it did not lodge such a claim at the time — it could have asked for a continuance to muster needed evidence. In the absence of such a request, it cannot now complain of unfair surprise. See generally United States v. Diaz-Villafane, 874 F.2d 43, 47 (1st Cir.1989) (explaining that it is “incumbent upon a party faced with [a surprise] situation to ask explicitly that the court grant the time needed to regroup, or waive the point”); Szeliga v. General Motors Corp., 728 F.2d 566, 568 (1st Cir.1984) (reciting that “the remedy for coping with surprise is not to seek reversal after an unfavorable verdict, but a request for continuance at the time the surprise occurs”).
. The government makes much of the fact that, during this interval, Suau recommended refinancing the farm loan by capitalizing the outstanding interest — a gambit that sometimes is regarded as a “mortal sin” in banking circles. In testimony before the district court, however, Somohano characterized Suau's proposal as a "last ditch effort” to avert foreclosure on a commercial loan, explaining that foreclosure is a last resort because it so frequently yields disappointing results. The government slruthiously ignores this uncontradicted testimony.
. Although our dissenting colleague glosses over it, this is the basis on which the district court required the U.S. Attorney, Guillermo Gil, and the RTC official, Bernard M. Brodsky, to testify at the hearing.
. We would take quite a different view if the judge's husband were a material witness in the recusal proceeding and had been called to testify as to disputed matters. Here, however, the government made no such argument and no one at any time indicated a wish to call Mr. Cerezo as a witness. The judge, therefore, was not placed in the position of passing upon her husband’s credibility.
. In this instance, our decision to set a lower threshold for the government, see supra at 30-31, makes no practical difference; if denying the recusal motion does not constitute an abuse of *36discretion, then, a fortiori, the government can show no clear entitlement to mandamus relief.