dissenting in part and concurring in part:
This appeal involves a flagrant instance of government misconduct in which the detective handling Calvin Dyess’s drug investigation seduced his wife, Ursala Dyess, and encouraged her to lie about critical *244facts regarding Calvin and his co-defendants’ culpability during the consolidated sentencing hearing. Further still, the detective permitted Ursala to keep some of the drug proceeds retrieved from Calvin’s backyard. In addition to this wrongdoing, another witness admitted she lied during the grand jury hearing, while still another witness testified during the sentencing hearing, recanted his testimony, and finally recanted his recantation.
These revelations came to light while the defendants’ sentences were on direct appeal to this Court. After we remanded the matter to Judge Haden, the district court judge who had originally sentenced defendants, to conduct further proceedings in light of this misconduct, the Supreme Court handed down its decision in United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2004). However, Judge Haden passed away before fully reconsidering the sentences, and the matter was transferred to Judge Faber.
Despite the obviously sensitive nature of these proceedings and Booker’s instruction to consider the Guidelines in connection with the sentencing factors set forth in 18 U.S.C. § 3553(a), id. at 259, 125 S.Ct. 738, Judge Faber did not conduct a de novo sentencing hearing. Instead, he conducted a limited sentencing hearing designed solely to address the degree to which Ursala Dyess and Detective Hart had tainted the other witnesses.
Despite the perjury that occurred at the initial sentencing hearing and the Booker errors arising from the original sentences imposed, the majority nonetheless concludes that the defects were harmless in light of the district court judge’s subsequent findings of fact. I disagree. . Judge Faber’s sentencing hearing could not and did not cure the errors. Thus, I dissent from the majority’s decision to affirm the sentences, as set forth in Parts VII and VIII of the opinion.
The district court judge explicitly stated that the sole purpose for the evidentiary hearing following remand from this Court “was to determine whether the alleged misconduct by Hart, Henderson, Rader [Ursula Dyess], or the purported recantations of Green and Cummings undermine the factual basis for the original sentencing guideline calculations in the defendants’ cases.” S.J.A. 381. Following the hearing, the judge concluded that: (1) any deficiencies in Ursula Dyess’s testimony were harmless in light of other testimony supporting well over 1.5 kilograms of cocaine activity by Calvin Dyess; (2) there was no evidence that Detective Hart compromised or tainted the testimony of witnesses other than Ursala Dyess at the original sentencings; (3) Green’s recantation of his testimony at the sentencing hearing was later recanted; and (4) Cummings’s perjury at the grand jury proceeding was harmless because she did not testify at the sentencing hearing. Thus, the district court judge concluded that “in spite of the credibility problems of Rader, Hart, Henderson, and Green, the original sentences are supported by abundant credible evidence.” S.J.A. 390.
The central flaw underlying this conclusion is that the district court judge was making after-the-fact credibility determinations. Instead, as the trier of fact, the judge on remand was obligated to make credibility determinations in the first instance. Here the judge considered solely the extent to which Ursala Dyess, Lori Nicole Cummings, Benjamin Green, Trooper Henderson,' and Detective Hart had perjured themselves, without considering the critical question of whether the testimony proffered by the remaining witnesses regarding the attributable drug *245quantities and relevant conduct for each defendant was credible.
Importantly, saying that the perjury of a few witnesses did not taint the remaining witnesses is not the same thing as saying that the remaining witnesses were credible. By attempting to rehabilitate Judge Haden’s findings by relying on a transcript of the sentencing hearing' — without the benefit of seeing or hearing the key witnesses — the district court judge essentially acted as an appellate court engaging in a sufficiency of the evidence analysis. See United States v. Burgos, 94 F.3d 849, 863 n. 5 (4th Cir.1996) (“[T]he appellate function is not to determine whether the reviewing court is convinced of guilt beyond reasonable doubt, but, viewing the evidence and the reasonable inferences to be drawn therefrom in the light most favorable to the Government, ‘whether the evidence adduced at trial could support any rational determination of guilt beyond a reasonable doubt.’ ” (internal citations omitted)). As the sentencing judge, a district court judge is in the best position to assess credibility, observe the demeanor of witnesses, resolve conflicting evidence, and determine the weight of the evidence. As appellate court judges, we necessarily depend on sentencing judges to make such fact-based determinations.
By not making these findings, the district court judge could not determine whether, after setting aside the tainted testimony, the evidence supported the drug quantities found and enhancements imposed by a preponderance of the evidence. On remand the district court judge was not entitled to rely on Judge Haden’s generalized findings of fact, which did not identify and evaluate individually the various witnesses who testified at the sentencing hearing. Judge Haden did not know at the time of sentencing that some witnesses had perjured themselves, and he made no specific credibility findings regarding the remaining witnesses. Judge Faber did not see the testimony of the remaining witnesses given and had no basis (beyond that which we, too, have as an appellate court) to evaluate its credibility. As a result, we have no way of knowing whether the testimony proffered by the remaining witnesses, untainted by Detective Hart’s egregious misconduct, could meet the government’s burden with respect to the challenged drug quantities and enhancements. It is ■ quite likely that Judge Haden’s findings of fact rested on the perjured testimony of Ursala Dyess and the questionable testimony proffered by Detective Hart, Trooper Henderson, and Benjamin Green. The government has failed to show that those findings of fact — and the sentences they justified' — • safely can rest on the testimony of the remaining witnesses instead.
Furthermore, the limited nature of the district court judge’s sentencing hearing foreclosed proper consideration of the sentencing factors set forth in 18 U.S.C. § 3553(a). Reconsidering the sentences following remand from this Court and Judge Haden’s death, the district court judge stated:
The court delayed ruling on these motions pending decision by the United States Supreme Court of [sic] United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005). The court has reconsidered Judge Haden’s sentences in these cases in light of Booker and finds them to be reasonable and fully supported by credible evidence. In light of this evidence, if called upon to resentence these defendants, the court would give them the same sentences Judge Haden imposed.
S.J.A. 390-91 (emphasis added).
Although a district court’s consideration of an alternative sentence under Booker *246could normally cure a Booker error, the district court judge’s curtailed hearing could not and did not cure the Booker errors in Judge Haden’s original sentences.* Given the limited purpose of the sentencing hearing on remand, the district court judge did not hear any evidence regarding the factors set forth in § 3553(a). As a result, he could not have made any of the factual findings required under Booker, such as: (1) the nature and circumstances of the offense and the history and characteristics of the defendant; (2) the need for the sentence imposed to reflect the seriousness of the offense, to promote respect for the law, and to provide just punishment for the offense; and (3) the need for the sentence imposed to provide the defendant with needed educational or vocational training, medical care, or other correctional treatment in the most effective manner. 18 U.S.C. § 3553(a); see United States v. Green, 436 F.3d 449, 455 (4th Cir.2006) (“[T]he court must make factual findings, as appropriate or necessary to carry out its sentencing function, and in every ease give the reasons for the sentence imposed, as well as reasons for particular deviations from the Sentencing Guidelines.” (citing 18 U.S.C. § 3553(c))). Thus, without hearing any evidence relevant to sentencing and making his own findings by a preponderance of the evidence, the district court could not fulfill its duty to consider the § 3553(a) factors in accordance with our post-Booker precedent. See United States v. Hughes, 401 F.3d 540, 546 (4th Cir.2005) (“[T]he court shall consider [the Guidelines] range as well as other relevant factors set forth in the guidelines and those factors set forth in § 3553(a) before imposing the sentence.”). Because the district court judge did not implement the jiost-Booker remedy by fully considering the § 3553(a) factors, the Booker errors stemming from the sentences cannot be deemed harmless.
As a result, the sentences suffer from two significant defects: (1) the strong likelihood that the underlying findings of fact rested on tainted testimony; and (2) Sixth Amendment error that was not remedied by the district court’s incomplete consideration of the factors set forth in § 3553(a). A de novo sentencing hearing could have eliminated the overwhelming appearance of impropriety that unfortunately has hung over these proceedings. Now we are left with sentences that may well have rested on tainted evidence.
For these reasons, I cannot join the majority in affirming these sentences. I would vacate the sentences and remand for resentencing.
As the majority acknowledges, Orange Dyess did not admit to any specific drug quantities. Nevertheless, Orange Dyess was subject to a base offense level of 36 and a 2-level enhancement for the presence of a firearm during the offense and received a sentence of 235 months. Without judicial factfinding, Orange Dyess would be subject to a total offense level of 6 and criminal history category of I — a sentencing range of 0-6 months, which is far less than the 235 months he received.
Similarly, Eric Spencer received a 3-level enhancement for his leadership role in the offense. Without the enhancement, he would have received a sentencing range of 188-235 months, which was less than the 262 months he received.
Finally, Calvin Dyess received a 2-level enhancement for the presence of firearms, a 4-level enhancement for his leadership role in the offense, and a 2-level enhancement for obstruction of justice. Without those enhancements, Calvin Dyess would have been subject to a sentencing range of 210-262 months, which is less than the life imprisonment he received.