Affirmed by published opinion. Judge MOTZ wrote the opinion, in which Judge DAVIS joined. Judge GREGORY wrote a separate opinion concurring in the judgment.
DIANA GRIBBON MOTZ, Circuit Judge:A jury convicted Dr. Erik Dehlinger of three counts of filing false income tax returns. He received a sentence of forty-two months imprisonment and one year of supervised release and was ordered to pay $363,207 in restitution and a fine of $5,000. Dehlinger appealed his conviction and sentence, and we affirmed. Dehlinger then moved for habeas relief, asserting that his trial counsel had labored under a prejudicial conflict of interest in violation of Deh-linger’s Sixth Amendment rights. Following an extensive evidentiary hearing, in which Dehlinger, his trial counsel, and other witnesses testified, the district court, in a thorough and well-reasoned opinion, denied Dehlinger habeas relief. The court did, however, grant Dehlinger a certificate of appealability pursuant to 28 U.S.C. § 2253. For the reasons below, we affirm the judgment of the district court.
I.
Dehlinger’s Sixth Amendment challenge rests on his trial counsel’s relationships with three individuals — Tara LaGrand, Gary Kuzel, and Collis Redd — who were involved in the same fraudulent scheme that gave rise to his convictions. Dehling*318er maintains that these relationships produced conflicts of interest that prevented his trial counsel from calling these individuals as witnesses to provide exculpatory testimony at his trial.
A.
Dehlinger’s convictions arose from his involvement with Anderson’s Ark and Associates (“AAA”), which marketed programs enabling users to avoid current income tax liability and “recapture” taxes paid in the previous two years. Dehlinger began using the AAA tax programs in 1999. He first became involved with the AAA through George Benoit, an employee of an AAA affiliate called Guardian Management, and Richard Marks, an AAA “planner,” i.e., an AAA employee who prepared client tax returns and other documents that formed the basis of the fraudulent tax schemes. Benoit prepared Dehlinger’s 1998, 1999, and 2000 tax returns using AAA’s tax schemes. Tara LaGrand, another AAA planner, prepared Dehlinger’s 2001 and amended 2000 tax returns. Use of the AAA programs resulted in a substantial benefit to Dehlinger. In the three years he used the programs, he avoided $363,207 in tax liability and obtained annual refunds on his income taxes despite earning, as an emergency room doctor, between $250,000 and $300,000 per year.
In 2002, the Government began its investigation of Dehlinger. The Government offered Dehlinger a plea agreement, in which he would plead guilty to one felony and cooperate with the Government. During plea negotiations, Robert Stientjes and other lawyers represented Dehlinger. When Dehlinger rejected the plea, the Government indicted him in August 2006. Then, Dehlinger, relying on a recommendation from one of Stientjes’s partners, retained Scott Engelhard as his trial counsel to work along with Stientjes at trial.
Dehlinger retained Engelhard based largely on Engelhard’s relative success as court-appointed counsel for AAA planner Tara LaGrand in her 2004 trial in Seattle, Washington.1 The jury deadlocked over the charges against LaGrand in that trial. Subsequently, LaGrand (still represented by Engelhard) accepted a guilty plea and was sentenced to twenty-four months imprisonment and one year of supervised release. In LaGrand’s plea agreement, she admitted that she knowingly prepared false loan statements and tax deductions. This directly contradicted her trial testimony, in which she had claimed that she did not know the AAA programs were illegal. LaGrand’s plea agreement contained a waiver of the right to appeal. Thus, Engelhard’s representation of LaG-rand at her trial effectively ended with her sentencing in September 2005.
One year after that representation ceased, and before undertaking his representation of Dehlinger, Engelhard obtained a conflict waiver from LaGrand. In that waiver, LaGrand identified Engelhard as her “former attorney” and stated that “[t]o the extent that there might be any apparent conflict of interest, I do hereby waive that conflict of interest so that Mr. Engelhard can represent Mr. Dehlinger at trial.” Accordingly, when retained by Dehlinger in the autumn of 2006, Engel-hard no longer represented LaGrand, and LaGrand had waived any continuing duties Engelhard might have owed her.
With Engelhard as his lead counsel, Dehlinger proceeded to trial in 2007. On *319October 15, 2007, after a four-day trial, the jury found Dehlinger guilty of tax fraud on his 1999, 2000, and 2001 tax returns.
A few weeks after Dehlinger’s conviction, Engelhard contacted LaGrand to inform her that the lawyers for other AAA clients had written him asking for her contact information. Those lawyers ultimately subpoenaed LaGrand to testify before the district court in Seattle, Washington in the prosecution of other AAA clients. Engelhard was appointed to serve as LaGrand’s counsel in this matter due to his familiarity with the AAA prosecutions. The district court in Seattle docketed En-gelhard’s re-appointment as LaGrand’s counsel as of November 29, 2007 — one and one half months following the conclusion of Dehlinger’s trial. Engelhard filed a motion to quash the subpoena of LaGrand on Fifth Amendment grounds. In March 2008, Dehlinger fired Engelhard when he learned that Engelhard had filed this motion on behalf of LaGrand.
B.
After his unsuccessful appeal of his convictions and sentence, Engelhard moved for habeas relief pursuant to 28 U.S.C. § 2255. The district court held a two-day evidentiary hearing on Dehlinger’s § 2255 motion. During that hearing, the court considered substantial documentary evidence and heard the testimony of several witnesses, including Dehlinger, Engelhard, Stientjes, and LaGrand.
Dehlinger testified that his sole defense at trial was his good faith reliance on the assurances of AAA planners that the AAA tax plans were legal. He asserted that he wanted Engelhard to call AAA planners LaGrand, Kuzel, and Redd to testify as to these assurances, but that Engelhard repeatedly advised him that their testimony would be harmful rather than helpful. Dehlinger contended that Engelhard’s decision not to call LaGrand, Kuzel, or Redd as witnesses was driven by a conflict of interest arising from Engelhard’s prior representation of them.
Dehlinger also offered evidence that LaGrand had written a novel based on Engelhard’s earlier representation of her, in which she depicted “Mr. Scott” (the character modeled after Engelhard) as a hero. LaGrand herself testified at the § 2255 hearing that she was “in awe of’ Engelhard. In addition, she stated that she believed Dehlinger’s tax returns were legal when she prepared them for Dehlinger and that she told him that. But LaG-rand acknowledged that she had ultimately pled guilty to fraud in connection with her preparation of AAA returns, invoked her Fifth Amendment privilege and refused to testify at the trial of another AAA defendant, and had not appeared voluntarily at the § 2255 hearing. Although Dehlinger argued that Engelhard represented LaG-rand during Dehlinger’s trial, he provided no evidence to support this contention.2
Engelhard testified that he “felt pretty clear ... before [he] even started to represent Dr. Dehlinger” that calling AAA planners as witnesses would not be the best way to present Dehlinger’s defense. In their preliminary discussions, Engelhard informed Dehlinger of this assessment. Nevertheless, Engelhard explored the option of calling AAA planners. He ultimately concluded that — given their status as convicted felons or affiliates of AAA, *320which a jury had found to be a fraudulent scheme — the risks inherent in their testimony outweighed any benefits.
As to Redd and Kuzel, Engelhard testified that Dehlinger said he had never interacted with them, and for this reason Engelhard concluded that they were not qualified to testify on Dehlinger’s behalf. As to LaGrand, Engelhard testified that he believed she would make an especially poor witness for Dehlinger because she had testified at her trial “that ■ she was completely innocent but then later entered a guilty plea” and so could be impeached on cross-examination. Engelhard also thought LaGrand’s testimony would not aid Dehlinger because, since she lacked experience in corporate and offshore matters, “a good prosecutor could make her look like someone that nobody should believe as a planner in complicated matters like this.” Co-counsel. Stientjes agreed with this assessment when Engelhard explained his reasoning to Stientjes and Deh-linger prior to undertaking Dehlinger’s representation. -
Engelhard also opted not to call as witnesses other AAA planners with whom he had no past relationship. Engelhard testified that he had considered including AAA affiliate Benoit as a witness. He believed that Benoit would make a better witness than LaGrand because the Government had not prosecuted Benoit, who had prepared two of the three tax returns that formed the basis of the charges against Dehlinger.. However, after meeting with Benoit, discussing the matter with co-counsel Stientjes, and conferring with an attorney representing other AAA clients, Engelhard concluded that the risks of Be-noit’s testimony outweighed the benefits. He reasoned, in an e-mail he sent to Deh-linger weeks before his trial, that “[i]f Benoit testified for us, all he could do is reiterate the points I can make without him. If he testifies and does not appear credible, then a jury might conclude that it was not reasonable for you to rely upon him.” Engelhard also decided against calling AAA planner Marks as a witness because, in addition to Marks’s conviction, Engelhard was familiar with Marks from the 2004 Seattle trial and found him to be a difficult personality. In contemporaneous e-mails, Engelhard explained all of these considerations to Dehlinger.
Instead of relying on AAA planners, En-gelhard determined that the best defense strategy was to rely on three other witnesses to establish Dehlinger’s defense: Scott Stringer, Bruce Burner, and Carl Chariot. Stringer, an expert, could and did opine that the structure of the AAA tax schemes was legitimate. Engelhard believed that “because of [Stringer’s] credentials and his lack of involvement with AAA programs, his opinion about the legality of the AAA programs would carry more weight with a jury than the testimony of an AAA Planner.”
Burner was a former co-worker of Deh-linger who also joined AAA. Engelhard believed that he would be a strong witness because “he relied upon the exact same information from AAA as Dr. Dehlinger for his belief that the AAA program was legal” and Burner’s “credibility was particularly strong because he had testified as a Government witness/victim in the Seattle trial.” Engelhard reasoned that “the Government would have no explanation for how they could treat Dr. Burner as a victim and yet treat Dr. Dehlinger as a criminal, and that the jury would feel compelled to acquit Dr. Dehlinger in light of Dr. Burner’s testimony.”
Engelhard further determined that Chariot, a prior associate of an IRS agent on whose assurances Dehlinger had assert-edly relied in becoming involved with AAA, would also make a strong witness. Char-*321lot could tell “a very compelling story” about how the agent “had conned [Chariot] into believing that the AAA program was legal.” Engelhard noted that Chariot had been a convincing witness at LaGrand’s trial in Seattle.
On May 16, 2012, the district court issued a thorough opinion denying Dehlinger habeas relief. The court found that Engelhard’s testimony was consistent with contemporaneous documentation and both “persuasive and credible.”3 In the court’s view, Engelhard’s decision not to call LaG-rand, Kuzel, or Redd as witnesses at Deh-linger’s trial “was based on trial strategy alone and not linked to any potential conflict of interest.”
Thus, the court concluded that Engel-hard’s “decision not to call witnesses who would be subject to impeachment for their prior crimes (LaGrand and Kuzel) or for their participation in a criminal organization (Redd) was simply a reasonable strategic decision which the record does not indicate that Dehlinger opposed.” The district court found that these decisions were “part of an overarching trial strategy, which [Engelhard] thought would be most beneficial to [Dehlinger],” and that “the record supported] the conclusion that Dehlinger and Stientjes acquiesced in this trial strategy.” The court further found that “this trial strategy was applied evenhandedly to ... other [AAA] witnesses [with whom Engelhard had no history of representation] which were not called for similar strategic reasons.”
Dehlinger noted a timely appeal.
II.
On appeal, Dehlinger again contends that- Engelhard provided ineffective representation that prejudiced him in violation of his Sixth Amendment rights. Sixth Amendment protections apply to every aspect of a lawyer’s representation. Deh-linger’s claim, however, rests entirely on one portion of Engelhard’s representation. Dehlinger contends that, in preparation for and during his trial, Engelhard labored under an active conflict of interest because of Engelhard’s history with LaGrand, Kuz-el, and Redd. Dehlinger maintains that this conflict adversely affected Engelhard’s performance because he should have, but did not, call them as witnesses.
Usually, a defendant can establish a Sixth Amendment ineffective assistance claim only by proof that (1) “counsel’s performance was deficient” and (2) “the deficient performance prejudiced the defense.” Strickland v. Washington, 466 U.S. 668, 687, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984).4 A lawyer’s “concurrent repre*322sentation” of multiple clients, however, raises a “high probability of prejudice.” Mickens v. Taylor, 535 U.S. 162, 175, 122 S.Ct. 1237, 152 L.Ed.2d 291 (2002). Accordingly, the Supreme Court has held that in concurrent representation cases, a defendant can establish a Sixth Amendment violation by “showing ... defective performance, but not requiring in addition (as Strickland does in other ineffectiveness-of-counsel cases), a showing of probable effect upon the outcome of trial.” Id. at 174, 122 S.Ct. 1237.
Arguably at least, this case concerns not concurrent but successive representation of assertedly conflicted clients. See Moss v. United States, 323 F.3d 445, 456, n. 15, 459 (6th Cir.2003) (distinguishing between concurrent representation, which is “the simultaneous representation of two or more co-defendants by [a] single attorney” and “[successive representation,” which occurs when a defendant’s counsel “has previously represented a co-defendant or trial witness”). The Supreme Court has specifically reserved the question of whether the second requirement of Strickland — namely proof of a prejudicial effect on the outcome of trial — applies to asserted conflicts arising from successive representation. See Mickens, 535 U.S. at 175-76, 122 S.Ct. 1237. We need not determine, however, whether Engelhard’s representation was successive and, if so, whether Dehlinger had to demonstrate that the asserted conflict had a probable prejudicial effect on the outcome of his trial. This is so because Dehlinger has failed to make the threshold showing that constitutionally deficient.
In order to establish constitutionally deficient performance on the basis of an alleged conflict of interest, a defendant, like Dehlinger, who has raised no objection at trial, must establish that (1) “an actual conflict of interest” (2) “adversely affected his lawyer’s performance.” Cuyler v. Sullivan, 446 U.S. 335, 348, 100 S.Ct. 1708, 64 L.Ed.2d 333 (1980). An actual conflict, which requires a defendant to show that his counsel “actively represented conflicting interests,” is the “constitutional predicate” for an ineffective assistance claim. Id. at 350, 100 S.Ct. 1708. Because “a possible conflict inheres in almost every instance of multiple representation,” id. at 348, 100 S.Ct. 1708, the mere “possibility of conflict is insufficient to impugn a criminal conviction,” id. at 350, 100 S.Ct. 1708.
But even when an “actual conflict” is shown, “an adverse effect is not presumed.” United States v. Nicholson, 475 F.3d 241, 249 (4th Cir.2007) (“Nicholson I ”).Instead, the defendant must separately prove that the conflict adversely affected his counsel’s performance by satisfying the three-prong test set forth in Mickens v. Taylor:
First, the [defendant] must identify a plausible alternative defense strategy or tactic that his defense counsel might have pursued. Second, the [defendant] must show that the alternative strategy or tactic was objectively reasonable under the facts of the case known to the attorney at the time of the attorney’s tactical decision.... Finally, the [defendant] must establish that the defense counsel’s failure to pursue that strategy or tactic was linked to the actual conflict.
240 F.3d 348, 361 (4th Cir.2001) (en banc) (citation omitted), aff'd without consideration of this point 535 U.S. 162, 122 S.Ct. 1237, 152 L.Ed.2d 291 (2002). The requirements for establishing an actual conflict and an adverse effect on the lawyer’s performance “are often intertwined, mak*323ing the factual analyses of them overlap.” United States v. Tatum, 943 F.2d 370, 375 (4th Cir.1991). The defendant must make both showings to obtain relief. Sullivan, 446 U.S. at 349-50, 100 S.Ct. 1708.
“Conflicts claims present mixed questions of law and fact that we review de novo.” Mickens, 240 F.3d at 360 (internal quotation marks omitted). “When the [district] court [has] conducted an evidentiary hearing prior to ruling, we review its findings of fact for clear error.” United States v. Nicholson, 611 F.3d 191, 205 (4th Cir.2010) (“Nicholson II”). “Because much of the adverse effect inquiry is heavily fact dependent, we believe appropriate deference should be given to the findings of the district court.” Mickens, 240 F.3d at 360. ‘When findings are based on determinations regarding the credibility of witnesses,” as they were in the case at hand, “we give even greater deference to the trial court’s findings.” United States v. Hall, 664 F.3d 456, 462 (4th Cir.2012) (internal quotation marks omitted).
With these principles in mind, we turn to the question of whether the district court erred in finding Dehlinger did not establish that a conflict of interest rendered Engelhard’s representation constitutionally deficient.
III.
A.
The district court based its denial of Dehlinger’s request for § 2255 relief on its determination that — even assuming ar-guendo that an actual conflict of interest existed — Dehlinger failed to satisfy the third prong of Mickens. That is, the district court focused on whether Engelhard’s decision to call witnesses other than LaG-rand, Kuzel, and Redd was “linked” to the asserted conflict, or instead was the product of a legitimate trial strategy. We too will frame our discussion around this third prong.
Dehlinger, as the defendant, bears the burden of proving the requisite “link.” Mickens, 240 F.3d at 361. To satisfy this burden, he must show Engelhard’s decision was not objectively reasonable. See Stephens v. Branker, 570 F.3d 198, 212 (4th Cir.2009)(finding that because the defendant failed to establish a link between an alleged conflict and counsel’s failure to pursue a defense the defendant “fail[ed] to demonstrate that the trial decisions made by his counsel were anything other than tactical judgments”). “If a reasonable attorney would have adopted the same trial strategy absent a conflict, [a defendant] cannot show [his lawyer’s] performance was adversely affected by that conflict.” Caban v. United States, 281 F.3d 778, 786 (8th Cir.2002).
B.
The district court carefully reviewed the evidence and made critical credibility determinations. After doing so, the court found that Engelhard’s decision not to call LaGrand, Kuzel, or Redd as witnesses amounted to nothing more than a reasonable strategic decision, which he made in good faith at the outset, applied evenhandedly to all potential witnesses, and adhered to consistently throughout trial. Ample evidence supports this finding.
With regard to Kuzel and Redd, Deh-linger had told Engelhard that he had no meaningful interaction with them. In fact, he barely knew them and they had not prepared any of his tax returns. Thus, it seems unlikely that either of them would have been permitted to testify as to Deh-linger’s assertedly innocent state of mind. Even assuming that they would have been permitted to testify, Dehlinger has failed *324to demonstrate that they could have provided testimony not elicited from the defense witnesses who did testify on his behalf at trial. See Eisemann v. Herbert, 401 F.3d 102, 108 (2d Cir.2005) (finding no adverse effect because nothing in the record “provide[d] the slightest indication as to what [the client-witness] would have said if called or even that he would have said anything at all”).
As to LaGrand, who is the primary focus of Dehlinger’s challenge, the record offers substantial support for the district court’s finding that Engelhard’s decision not to call her was driven by strategic concerns rather than any conflict. The testimony and affidavits from Engelhard, Stientjes, and Dehlinger himself, as well as contemporaneous e-mail communications, establish that Engelhard and co-counsel Stientjes believed that the risks of LaG-rand’s testimony outweighed the benefits, and informed Dehlinger of this strategy. For instance, Dehlinger e-mailed Engel-hard and Stientjes on September 20, 2007 (three weeks prior to trial) to ask about the witness list. After inquiring about other potential witnesses, Dehlinger asked Engelhard and Stientjes whether they “still think that Mar[ks] and/or LaGrand would be too risky.” (emphasis added). Stientjes responded to Dehlinger’s e-mail informing him that “[Engelhard] and I have talked about the benefits [and] burdens of each witness.... I think we have a great defense with our core witnesses. I think we only have risk in calling any more.”
This evidence thus belies Dehlinger’s claim that Engelhard led him to believe up until the eve of trial that LaGrand would be called as a witness. Instead, the evidence establishes just the opposite. First, Engelhard told Dehlinger prior to undertaking his representation that, in his expert opinion, LaGrand’s testimony would not help Dehlinger, and Dehlinger chose to retain Engelhard with this knowledge.5 Then, Engelhard and co-counsel adhered to this strategy and kept Dehlinger apprised of it throughout trial preparations.
As the district court found, the strategic considerations that led Engelhard to conclude that calling LaGrand would be “too risky” were objectively reasonable. Not only was LaGrand a convicted felon; she had also pled guilty to tax fraud involving the very same fraudulent organization that provided the basis for the charges against Dehlinger. Moreover, LaGrand had a history of conflicting statements under oath about this tax fraud. This conflict made it impossible to predict what she would say on the stand and rendered any testimony from her vulnerable to blistering cross-examination. Furthermore, because LaG-rand played no part in Dehlinger’s decision to follow the AAA tax “plan” and did not prepare most of the tax returns that formed the basis for Dehlinger’s indictment, it is difficult to see that anything *325helpful to Dehlinger would be gained from her testimony that was not obtained from the more reliable witnesses Engelhard used to establish Dehlinger’s defense.
Even more than Kuzel and Redd, LaG-rand was a witness with very little upside and a substantial downside — involvement in the same tax fraud as Dehlinger and, in her case, initial denial of her involvement and then admission of and imprisonment for it. No adverse effect results from a trial lawyer’s decision not to call witnesses whose testimony would be cumulative or potentially damaging to a defendant’s case. See Winfield v. Roper, 460 F.3d 1026, 1033-34 (8th Cir.2006) (citing cases).6
Dehlinger’s contentions that LaGrand, Kuzel, or Redd would have provided exculpatory testimony at his trial rest on nothing more than conjecture. One can only speculate as to what they would have said and what effect this testimony might have had on the jury. Evaluation of the testimony of possible witnesses is precisely the sort of strategic decision entrusted to the professional judgment of trial counsel. The record provides abundant evidence that, as the district court found, Engel-hard’s decision not to call LaGrand, Kuzel, Redd or any other AAA planner was an objectively reasonable one, which was based on Engelhard’s familiarity with the facts of the case and his thorough investigation of the best options available to his client.
IV.
The Sixth Amendment does not provide a basis for disappointed clients to launch after-the-fact attacks on the objectively reasonable strategic decisions of their trial attorneys. The district court did not err in finding that Dehlinger failed to establish that Engelhard’s representation was anything other than objectively reasonable. For the foregoing reasons, the judgment of the district court is
AFFIRMED.
. In 2002, Engelhard also engaged in preliminary representation discussions with two other AAA planners — Gary Kuzel and Collis Redd — but later advised both of them that he would represent only LaGrand.
. Instead, the evidence produced at the § 2255 hearing established that Engelhard’s representation of LaGrand in her criminal proceedings ended in September 2005, more than one year before he joined Dehlinger's defense team; Engelhard's appointment as LaGrand’s counsel to file the motion to quash the subpoena in the separate proceeding occurred on November 29, 2007, several weeks after Dehlinger’s trial had concluded.
. The court also made credibility determinations with regard to the testimony of co-counsel Stientjes and Dehlinger's expert witness. Stientjes initially testified at the § 2255 hearing that “up until the last minute, Dr. Deh-linger expected Tara LaGrand ... to testify for Dr. Dehlinger/' but he admitted on cross-examination that he meant only that Dehlinger expected her to be available to testify if needed. The district court refused to credit Stientjes's testimony, which failed to withstand cross-examination and which also conflicted with Stientjes’s affidavits and contemporaneous e-mails. The court also gave little weight to the testimony of Dehlinger’s expert witness who relied uncritically upon the truth of Stientjes’s contradicted statement to reach an opinion about the ultimate legal issues before the court.
. In assessing whether counsel’s performance was constitutionally deficient, "[t]he purpose is simply to ensure that criminal defendants receive a fair trial,” Strickland, 466 U.S. at 689, 104 S.Ct. 2052, "not to enforce the Canons of Legal Ethics,” Mickens, 535 U.S. at 176, 122 S.Ct. 1237. Thus, although the rules of attorney professional conduct may be relevant as a guide to assessing reasonableness, “breach of an ethical standard does not necessarily make out a denial of the Sixth Amendment guarantee of the assistance of *322counsel.” Nix v. Whiteside, 475 U.S. 157, 165, 106 S.Ct. 988, 89 L.Ed.2d 123 (1986).
. Dehlinger acknowledged at the § 2255 hearing that prior to agreeing to represent him, Engelhard stated that he did not think it would be a good strategic decision to call LaGrand, and "that if that's the way [Dehlinger] wanted to go, [he] really should go with some other lawyer.” Dehlinger argues that this statement shows Engelhard placed LaG-rand’s interests above Dehlinger’s by threatening not to represent Dehlinger unless he agreed not to call LaGrand as a witness. Even if the statement could be construed as evidence of a conflict during representation, as Dehlinger maintains, it could also be construed as evidence that Engelhard simply provided his prospective client (Dehlinger) with the frank professional judgment that LaGrand would make a poor witness and that calling her would not be the best strategy for presenting Dehlinger’s defense. Given the deference we must give the district court’s findings, we cannot hold that the district court erred in so finding.
. The evidence produced at the § 2255 hearing also strongly suggests that LaGrand, even if called, would not have agreed to testify at Dehlinger’s trial. At the time of that trial, she was on supervised release and limitations had not run on charges the Government waived as part of her plea bargain. At her own trial, she had claimed total innocence; when pleading guilty, she admitted to knowingly engaging in fraud. Any further testimony exposed her to charges of perjury and breach of her plea agreement.