In the

    United States Court of Appeals
                 For the Seventh Circuit
                     ____________________
No. 19-1938
UNITED STATES OF AMERICA,
                                                   Plaintiff-Appellee,
                                 v.

EARL R. ORR,
                                               Defendant-Appellant.
                     ____________________

         Appeal from the United States District Court for the
                     Central District of Illinois.
           No. 2:16-cr-20052 — Sara Darrow, Chief Judge.
                     ____________________

     ARGUED MAY 21, 2020 — DECIDED AUGUST 10, 2020
                ____________________

   Before MANION, HAMILTON, and BRENNAN, Circuit Judges.
    BRENNAN, Circuit Judge. A search warrant for illegal drugs
at the home of Earl Orr led to his arrest for possessing a fire-
arm as a felon. After a two-day trial, a jury found him guilty.
Orr appeals a number of decisions made by the district court
before and during that trial.
  We conclude that the district court properly denied Orr’s
motion to suppress evidence. But Judge Bruce, who presided
2                                                 No. 19-1938

over this case at trial, had engaged in improper ex parte com-
munications with the U.S. Attorney’s Office in other matters.
That cast a pall over certain decisions in this case which re-
quired the exercise of substantial discretion. This was not
harmless error, so we vacate Orr’s conviction and remand for
further proceedings before a different judge.
                               I.
    In March 2016, a confidential source known as “Dave
Bonz” told a member of the Champaign Police Department
that he knew a crack cocaine dealer named Moe. Bonz had
provided police with information on drug dealers in the past
and had participated in three controlled buys. According to
Bonz, Moe had sold him crack cocaine on several occasions.
Each time, Bonz dialed a number ending in 1335 and Moe de-
livered the crack cocaine in a four-door maroon Mitsubishi
registered in Illinois to Moe’s girlfriend.
    Over a few months, officers with the department con-
ducted five controlled buys from Moe. Each time Bonz called
Moe at the number ending in 1335 and bought crack cocaine
from Moe or one of his associates using pre-recorded or
marked money. Officers surveilled all five of the controlled
buys, and three of the transactions were recorded with a cov-
ert video-recording device. During four of the controlled
buys, officers watched a maroon Mitsubishi described by
Bonz travel between the meet location and an apartment on
Smith Road in Urbana, Illinois.
    After reviewing the video footage, officers identified Moe
as Earl Orr, who was on parole after being convicted of un-
lawful possession of a controlled substance with intent to de-
liver. Orr’s identity was confirmed in three ways. First,
No. 19-1938                                                    3

officers showed Bonz a picture of Orr from a law enforcement
database, with all identifiers concealed. Bonz identified Orr as
Moe. Second, officers tied the maroon Mitsubishi to Orr. Of-
ficers discovered that the Mitsubishi was registered to Jakaeya
Biggers, and, after being presented with a copy of Biggers’
driver’s license, Bonz identified Biggers as “Moe’s girlfriend.”
Third, officers linked the apartment on Smith Road to Orr.
Both a law enforcement database and a list of tenants pro-
vided by the owner of the apartment revealed that Orr and
Biggers lived together at the apartment on Smith Road.
    The police filed for a search warrant of Orr’s apartment
and included an affidavit in which they described the infor-
mation provided by Bonz, the corroboration performed by of-
ficers, and the controlled buys. A judge issued the search war-
rant and officers searched Orr’s apartment. They found a .25
caliber semi-automatic pistol along with ammunition, ap-
proximately 22 grams of crack cocaine, approximately 15
grams of powdered cocaine, a digital scale, razor blades, and
five boxes containing small Ziplock baggies. After officers ar-
rested Orr and read him his Miranda rights, Orr voluntarily
admitted that the gun and cocaine were his, and he reaffirmed
ownership of those items during a second interview after the
search. A grand jury later charged Orr with possessing a fire-
arm as a felon in violation of 18 U.S.C. § 922(g), and his case
proceeded to a jury trial. Because the gun was not found in
Orr’s actual possession, the prosecution’s case centered on cir-
cumstantial evidence and Orr’s confessions.
   Before trial, Orr moved to suppress the evidence gathered
from his apartment, asserting Bonz was an unreliable source.
The district court denied Orr’s motion. It found Bonz reliable
and, in the alternative, that any defects in his credibility were
4                                                 No. 19-1938

remedied through the controlled buys and the good faith ex-
ception outlined in United States v. Leon, 

(1984).
    Orr and the government then moved in limine concerning
the admissibility of the drug evidence recovered during the
search and the controlled buys. The government argued the
drug evidence proved Orr’s motive for possessing the gun.
Orr disagreed, asserting the drug evidence was irrelevant to
the gun charge and unduly prejudicial. The district court
granted the government’s motion in limine and denied Orr’s
motion in limine. Even so, the district court conditioned the
admissibility of the drug evidence on whether Orr placed his
motive for possessing the gun at issue during trial. After Orr
took the stand and testified that he did not have any reason to
possess a firearm in response to a question asked by the gov-
ernment, the prosecutor and the defense attorney argued at
sidebar about whether Orr had placed his motive at issue. The
district court agreed with the government and ruled that Orr
had placed it “squarely [] at issue” by claiming he had no rea-
son to possess a firearm, so the prosecutor was allowed to pre-
sent evidence of Orr’s drug involvement. As a result of this
ruling, a witness for the prosecution testified that Orr had
drug dealing paraphernalia and “several thousand dollars[‘]
worth of drugs” stored in his apartment. Following the close
of evidence, the district court instructed the jury to consider
this testimony only in the context of whether Orr had a motive
to possess the gun.
    Also during trial the district court permitted the prosecu-
tor to cross-examine Orr on his prior conviction for unlawful
possession with intent to deliver a controlled substance, to
which Orr’s counsel did not object. On this topic the court
gave the jury another limiting instruction, directing them to
No. 19-1938                                                      5

consider evidence of Orr’s prior conviction only when evalu-
ating the credibility of his testimony and whether he was a
convicted felon at the time he was alleged to have possessed
the gun.
    The jury found Orr guilty. Before sentencing, the Judicial
Council of the Seventh Circuit determined that the trial judge,
Judge Bruce, had breached the Code of Conduct for U.S.
Judges by engaging in improper ex parte communications in
other cases with members of the U.S. Attorney’s Office for the
Central District of Illinois. In re Complaints Against Dist. Judge
Colin S. Bruce, Nos. 07-18-90053, 07-18-90067 (7th Cir. Jud.
Council May 14, 2019). Although the Judicial Council found
no evidence that those communications affected the outcome
of any case, the Council suspended Judge Bruce from all crim-
inal matters involving the U.S. Attorney’s Office for the Cen-
tral District of Illinois for one year.

was transferred to another judge for sentencing and Orr re-
ceived 210 months of imprisonment.
                                  II.
    Orr raises a number of issues on appeal. He argues the dis-
trict court erred by denying his pretrial motion to suppress
the evidence gathered from his apartment. He also submits
that the district judge should have recused himself because of
his ex parte communications with the U.S. Attorney’s office in
other cases. He further challenges the admission of drug evi-
dence under Federal Rule of Evidence 404(b) and the
6                                                   No. 19-1938

allowance of cross-examination questions about his prior fel-
ony conviction. We begin with the suppression ruling.
    A. The Motion to Suppress
    Orr contends the search warrant was not supported by
probable cause because Bonz was neither credible nor relia-
ble. Orr’s argument fails, however, because the affidavit es-
tablished probable cause and, in the alternative, the good faith
exception in Leon applies.
    Under the Fourth Amendment, warrants may not be is-
sued “but upon probable cause.” U.S. CONST. amend. IV.
Probable cause exists when, considering the totality of the cir-
cumstances, there is a “fair probability that contraband or ev-
idence of a crime will be found in a particular place.” Illinois
v. Gates, 

, 238 (1983). To determine whether an is-
suing judge correctly determined that probable cause for a
search existed, district courts must give “‘great deference’ to
the issuing judge’s determination so long as the judge had a
‘substantial basis’ for the finding.” United States v. Miller, 

, 692–93 (7th Cir. 2012) (citations omitted). This court
reviews the district court’s probable cause determination de
novo but, like the district court, must also give “‘great defer-
ence’ to the conclusion of the judge who initially issued the
warrant.” United States v. Searcy, 

, 1122 (7th Cir.
2011) (quoting United States v. Garcia, 

, 485 (7th
Cir. 2008)).
   When, as here, the information used to support a probable
cause finding is primarily derived from an informant’s tip,
“the legitimacy of [the] probable cause determination turns
on that ‘[informant]’s reliability, veracity and basis of
knowledge.’” United States v. Olson, 

, 370 (7th Cir.
No. 19-1938                                                     7

2005) (quoting United States v. Johnson, 

, 1038 (7th
Cir. 2002)). Courts assess an informant’s credibility by consid-
ering: “(1) the degree of police corroboration; (2) the inform-
ant’s firsthand knowledge; (3) the detail provided; (4) the time
between the reported events and the warrant application; and
(5) whether the informant appeared before the judge.” United
States v. Haynes, 

, 665 (7th Cir. 2018) (citing United
States v. Johnson, 

, 600 (7th Cir. 2011)). We review
how those factors were considered.
    First, the extensive police corroboration detailed in the af-
fidavit strongly supports the issuing judge’s probable cause
determination. Officers linked the maroon Mitsubishi and
apartment on Smith Road to Orr and his girlfriend. Officers
also determined that Orr had a prior conviction for dealing
drugs. See United States v. Bell, 

, 1053 (7th Cir.
2009) (deciding that a defendant’s prior conviction for a drug-
related crime helped establish probable cause for a drug-re-
lated search). Most importantly, however, Orr sold crack
cocaine to Bonz while under police surveillance. “Generally,
a controlled buy, when executed properly, is a reliable indica-
tor as to the presence of illegal drug activity.” United States v.
Sidwell, 

, 869 (7th Cir. 2006) (footnote omitted).
Here, officers conducted not one but five controlled buys in
the month before the search warrant’s execution. Over the
course of these controlled buys, officers watched Orr as he
sold substances to Bonz, confirmed those substances con-
tained cocaine base, and observed the maroon Mitsubishi
traveling between Orr’s apartment and the pre-arranged deal
locations. Even though Bonz did not report seeing crack co-
caine in Orr’s house, these facts are strong evidence that Orr
stored crack cocaine in his apartment. See 

(finding “a ‘fair probability’ that the [defendant’s] house
8                                                   No. 19-1938

contained evidence of illegal activity” after the defendant left
his house, sold crack cocaine to an informant, and then re-
turned to his house).
    Next, on the second and third factors, Bonz had firsthand
knowledge of Orr’s drug dealing, and he shared that
knowledge in detail with police. Nevertheless, Orr argues the
affidavit was deficient because Bonz never described Orr’s
identifying features, the quantity of drugs he believed was on
Orr’s person, or the quantity of drugs he believed was at Orr’s
residence. None of these arguments are persuasive. Although
Bonz first described Moe only as a Black male, that descrip-
tion is not problematic because Bonz later identified Moe as
Orr when presented with a picture of Orr from a law enforce-
ment database. Nor does the affidavit’s failure to mention the
quantity of cocaine possessed by Orr on his person or in his
residence pose a problem. Precedent does not require a confi-
dential informant to provide officers with every detail of illicit
conduct. See United States v. Garcia, 

, 485–86 (2008)
(concluding affidavit established probable cause despite fail-
ing to mention how much cocaine was seen by the informant).
Here, Bonz gave officers Orr’s telephone number and de-
scribed how Orr delivered cocaine. On these facts, the second
and third factors support a probable cause finding.
    The fourth factor concerns timing. The last controlled buy
occurred within days of the search. Our court has found sim-
ilar timeframes support probable cause findings under the
fourth factor. See, e.g., 

(“This infor-
mation was [] transmitted within a relatively short period of
time—72 hours—before the application for the search warrant
and certainly was not stale.”); 

(“The in-
formation here was fresh (3 days old).”). Therefore, the time
No. 19-1938                                                  9

between the last controlled buy and the search supports a
probable cause finding. But the fifth factor weighs against
probable cause. Bonz did not testify in front of the issuing
judge, depriving the judge of the opportunity “to evaluate the
informant's knowledge, demeanor, and sincerity.” United
States v. Koerth, 

, 866 (7th Cir. 2002).
    Because four of the five factors support a probable cause
determination, a reasonable fact finder could conclude that
the warrant affidavit set forth facts to establish probable
cause. And if the affidavit was deficient in some respect,
namely Bonz’s failure to testify, the controlled buys provide
strong enough corroboration to support a probable cause
finding. See 

(“A properly executed con-
trolled buy can establish probable cause, even when the tip
that prompted it might not have been reliable.”); United States
v. Brack, 

, 756 (7th Cir. 1999) (“[A] deficiency in
one factor may be compensated for by a strong showing in
another or by some other indication of reliability.”).
    Even if the affidavit for the search warrant failed to estab-
lish probable cause, the good faith exception in Leon provided
alternative grounds to reject Orr’s suppression motion. Under
the good faith exception, “[a] facially valid warrant issued by
a neutral, detached magistrate will be upheld if the police re-
lied on the warrant in good faith.” United States v. Peck, 

, 757 (7th Cir. 2003) (citing 

,
922-23). The district court found that Officer Cully Schewska,
who swore out the affidavit, relied on the search warrant in
good faith, and we review that determination de novo. See

.
   Officer Schewska’s decision to obtain a search warrant is
prima facie evidence of good faith. See 

10                                                No. 19-1938

n. 21. Orr may rebut this presumption by showing that the is-
suing judge “wholly abandoned his judicial rule” or that the
affidavit was “so lacking in indicia of probable cause as to ren-
der official belief in its existence entirely unreasonable.”

(quotation marks omitted) (citing 

, 923)). Orr’s sole argument here is that Officer
Schewska’s reliance on the warrant was unreasonable because
he omitted mention of Bonz’s history of criminal behavior and
substance abuse in the affidavit. But even if such omissions
were significant, the warrant affidavit contained extensive
corroboration, referenced detailed information gathered
firsthand by Bonz and Champaign police officers, and re-
ferred to a controlled buy that occurred only days before.
Given these details, no reasonable officer would have be-
lieved the search of Orr’s apartment was unconstitutional.
Therefore, the district court correctly rejected Orr’s argument
under Leon.
     B. The Judicial Recusal Statute
    Orr next argues he is entitled to a new trial because the
trial judge’s ex parte communications with the prosecuting
U.S. Attorney’s Office violated 28 U.S.C. § 455(a), the judicial
recusal statute. Under § 455(a), “[a]ny justice, judge, or mag-
istrate judge of the United States shall disqualify himself in
any proceeding in which his impartiality might reasonably be
questioned.’’
   Although the government concedes Judge Bruce’s
conduct violated this statute, it argues that any error was
harmless. “Not every violation of § 455(a) warrants a drastic
remedy, like a new trial.” United States v. Williams, 

, 1063 (7th Cir. 2020). Mere appearance of impropriety is
not enough for reversal and remand—a party must show a
No. 19-1938                                                      11

risk of harm. See

Bergeron, 

, 883 (7th
Cir. 2011)). To determine whether Judge Bruce’s violation is
harmless, we consider the three factors announced in Liljeberg
v. Health Servs. Acquisition Corp., 

(1988): (1) “the
risk of injustice to the parties in the particular case,” (2) “the
risk that the denial of relief will produce injustice in other
cases,” and (3) “the risk of undermining the public’s confi-
dence in the judicial process.”

see Williamson v.
Indiana Univ., 

, 464 (7th Cir. 2003) (applying the
Liljeberg factors to claim under § 455(a)).
    Before applying the Liljeberg factors, we provide more
background. This is not the first case to come before our court
arising out of Judge Bruce’s ex parte communications. In
United States v. Atwood, 

(7th Cir. 2019), our court
reviewed Judge Bruce’s sentencing of a defendant after he
pleaded guilty to federal drug crimes.

Atwood
argued that because of Judge Bruce’s ex parte communica-
tions in other cases, he was entitled to resentencing.

We vacated Atwood’s sentence and remanded his case for re-
sentencing by a different judge.

We ruled that all
three Liljeberg factors counseled remand “[b]ecause of [Judge
Bruce]’s broad discretion in sentencing.”

    In United States v. Williams, 

(7th Cir. 2020),
we decided whether Judge Bruce’s ex parte communications
in other cases entitled Williams, a criminal defendant, to a
new trial.

Judge Bruce presided over Williams’s
trial at which he was convicted, but the case was transferred
to another judge for sentencing.

Unlike the defend-
ant in Atwood, Williams did not show that Judge Bruce made
any decisions that involved broad discretion. All pre-trial and
trial rulings in Williams were “minimal” and “none [were]
12                                                 No. 19-1938

challenge[d] on 

. And because Wil-
liams was sentenced by another judge, he was unable to argue
that Judge Bruce exercised discretion in sentencing as in
Atwood. See 

(“Judge Bruce did not
preside over Williams’s sentencing hearing. This distinction
matters because judges generally have more discretion over
sentencing than the outcome of a jury trial.”). Although we
affirmed Williams’s conviction after finding that all three of
the Liljeberg factors suggested Judge Bruce’s § 455(a) violation
was harmless error, we clarified that the first and third
Liljeberg factors could have come out differently had Judge
Bruce issued discretionary rulings. 

-65.
    Like the defendant in Williams, Orr appeals his conviction
after a jury trial presided over by Judge Bruce. But unlike the
defendant in Williams, Orr challenges three seemingly discre-
tionary decisions by Judge Bruce: the denial of the motion to
suppress, the admission of drug evidence under Federal Rule
of Evidence 404(b), and the allowance of cross-examination
questions about Orr’s prior felony conviction.
    A few months before trial, Judge Bruce issued the order
denying Orr’s motion to suppress the evidence recovered
during the apartment search. While Orr contends Judge
Bruce’s suppression ruling was a close discretionary call, we
disagree. As discussed above, four of the five factors our court
uses to determine an informant’s credibility and reliability
strongly supported the district court’s probable cause finding.
The five controlled buys, all of which occurred under police
surveillance, provided persuasive evidence that a search of
Orr’s apartment would reveal controlled substances. Even
more, the good faith exception created in Leon furnishes a
No. 19-1938                                                  13

compelling and alternative rationale for denying Orr’s motion
to suppress. Given the manifest facts and applicable law,
Orr’s appeal of the district court’s suppression ruling fails to
present a colorable claim. Because no reasonable district court
would have reached a different result, the suppression ruling
here required little discretion and it does not affect our analy-
sis of this case under Liljeberg.
    But Orr’s challenges to two of Judge Bruce’s trial decisions
are a different matter. The first was the district court’s admis-
sion into evidence of drugs and drug paraphernalia gathered
during the controlled buys and search of Orr’s apartment.
When deciding this question before trial, this was not a diffi-
cult choice in the event Orr placed his motive at issue. This
court had already decided—in a similar case Judge Bruce re-
lied on—that such evidence is admissible under Federal Rule
of Evidence 404(b). See United States v. Schmitt, 

,
534 (7th Cir. 2014) (“[T]he evidence proffered by the govern-
ment was relevant to motive. The testimony that [the defend-
ant] was a drug dealer and that drugs were found in his home
when he was arrested was relevant to suggest to the jury why
he would have a firearm.”).
   But when Orr said he had no reason to own a firearm,
Judge Bruce made a close discretionary call by deciding that
Orr placed his motive at issue. While Orr was on the witness
stand the prosecutor asked: “You didn’t have any reason to
possess a firearm?” Orr responded: “I haven’t, I haven’t
touched a firearm in 25 years, sir.” “[B]ased upon [Orr’s] an-
swers, … tone[,] and manner,” the district court determined
Orr placed his motive at issue. Yet given this exchange,
whether Orr or the prosecutor placed motive at issue is not a
simple question. As Orr points out on appeal, he denied
14                                                             No. 19-1938

having a reason to possess a firearm in response to the prose-
cutor’s questioning. Orr asserts the prosecutor placed motive
at issue when he asked Orr if he had any reason for possessing
a firearm. The parties dispute who opened the door to admit-
ting the drug evidence, and this evidentiary ruling involved a
substantial amount of discretion.
    The second close discretionary call the district court faced
was when the prosecutor asked Orr if he had been convicted
of dealing drugs and if that conviction should affect his cred-
ibility. The district court permitted the prosecutor’s line of
questioning but cautioned him not to “get into a prejudicial
area” by “overplay[ing] it.” Although the parties on appeal
characterize the government’s inquiry as potentially falling
under the “motive” exception of Federal Rule of Evidence
404(b), the questioning likely occurred within the parameters
of the impeachment exception contained in Federal Rule of
Evidence 609. The district court issued a pre-trial order clari-
fying that Orr’s prior conviction could not be introduced to
prove motive, and the prosecutor mentioned Orr’s prior con-
viction only in the context of impeaching him. Further, the
jury was instructed to consider Orr’s prior conviction only
when deciding the credibility of his testimony and whether
he was a felon at the time he possessed the gun. All of these
facts indicate the district court permitted the prosecutor to im-
peach Orr under Rule 609. 1 But regardless of which rule the
questioning occurred under, the district court exercised sub-
stantial discretion by weighing the probative value and

     1 Because the district court only briefly addressed the prejudicial effect

of the prosecutor’s questioning, it is difficult to conclusively determine
whether the district court applied the stricter balancing test contained in
Rule 609(a)(1)(B) or more lenient balancing test in Rule 403.
No. 19-1938                                                      15

prejudicial effect of the questioning and by allowing the ques-
tioning to proceed.
    So two discretionary rulings distinguish this case from
Williams. The pre-trial and trial rulings in Williams were rou-
tine, granted in favor of both parties, uncontested on appeal,
and not overly prejudicial to the defendant. See 

. The two discretionary rulings in this case were
non-routine, decided in favor of the government, and
challenged on appeal. Notably, both rulings in this case sig-
nificantly aided the prosecution. In the first, the district court
permitted the prosecutor to introduce evidence that Orr
stored drug-dealing paraphernalia and “several thousand
dollars[‘] worth of drugs” in his apartment. As a result of the
second, Orr was not only impeached on his felony conviction
but the jury was presented with evidence that he was
convicted of dealing drugs. Because this case centered on cir-
cumstantial evidence and credibility determinations, both de-
cisions prejudiced Orr. With these discretionary decisions in
mind, we turn to the Liljeberg factors.
    The first Liljeberg factor requires us to consider “the risk of
injustice to the parties.” 

. We start with
the potential injustice Orr may suffer if we upheld his convic-
tion. The record suggests that upholding Orr’s conviction
would create a tangible risk of unfairness to him. Because of
the discretionary calls described above, it is possible the dis-
trict court’s personal biases influenced the outcome in this
case. See 

. For the first factor, though,
we must also consider the risk of injustice to the government
if a new trial is granted. Retrying this case would likely re-
quire the government to “spend valuable time and
money … thereby diverting resources from other cases.”
16                                                    No. 19-1938

. Even so, the risk of injustice to the
government is directly related to the complexity of the trial.
See United States v. Cerceda, 

, 815 (11th Cir. 1999)
(en banc) (per curiam) (“[T]he government would face great
hardship if forced to conduct a new trial [] because of the com-
plexity of the case (a 78 count, complex white-collar prosecu-
tion the trial of which lasted two-and-a-half months).”). We
conclude that the risk of injustice to the government in this
matter is relatively slight due to the straightforwardness and
brevity of the prosecution’s case. Orr faced one charge, and
the trial lasted only two days. On these facts, the risk of injus-
tice Orr faces if we do not vacate his conviction is greater than
the risk of injustice the government faces if we upheld Orr’s
conviction. So the first Liljeberg factor favors Orr.
    Under the second Liljeberg factor, we look to “the risk that
the denial of relief will produce injustice in other cases.”

. The parties in this case raise the same
arguments as in 

. The government
contends no further action is necessary to induce other judges
to exercise caution in their communications because Judge
Bruce was thoroughly investigated, those results were
adopted by the Judicial Council, he was publicly repri-
manded, and he has implemented new practices to prevent
similar issues in the future. Orr, on the other hand, argues
these facts are not enough to ensure judges exercise more cau-
tion in the future and that further action must be taken. In
Williams, we balanced these arguments and decided that the
second Liljeberg factor counsels against awarding 

. But see 

(finding the sec-
ond Liljeberg factor counsels in favor of resentencing). Because
no reason is provided as to why the Williams decision was
No. 19-1938                                                            17

erroneous on this point, we conclude this factor favors up-
holding Orr’s conviction.
    The third Liljeberg factor requires us to consider “the risk
of undermining the public’s confidence in the judicial pro-
cess.” 

. Like the defendant in Williams,
Orr was found guilty by a jury of his peers. Although in
Williams we decided that the jury finding the defendant guilty
was “significant,” we envisioned “a case where a judge has
substantial discretion and his rulings have a significant im-
pact on the outcome, thus undermining the public confidence
in the judicial 

. Such a case is now
before us. Judge Bruce exercised substantial discretion by ad-
mitting evidence of Orr’s drug dealing and by permitting the
prosecutor to cross-examine Orr on his felony conviction for
dealing drugs. These evidentiary decisions were particularly
consequential because they bolstered the prosecution’s case,
which rested on circumstantial evidence and credibility calls.
Given these discretionary rulings, upholding Orr’s conviction
may damage the public’s confidence in the impartiality of the
judiciary. For these reasons, the final Liljeberg factor favors va-
cating Orr’s conviction.
   The first and third Liljeberg factors support vacating Orr’s
conviction, so we cannot conclude the error in Judge Bruce
not disqualifying himself from the case was harmless. Ac-
cordingly, we vacate his conviction. 2



    2 Because we remand for a new trial, we need not address Orr’s other
two arguments that the district court erred by admitting the drug evidence
under Federal Rule of Evidence 404(b) and by allowing the prosecution to
cross-examine him on his prior felony conviction.
18                                               No. 19-1938

                              III.
   For these reasons, we AFFIRM the district court’s suppres-
sion ruling, VACATE Orr’s conviction and sentence, and
REMAND for further proceedings before a district judge other
than Judge Bruce.