McKAY, Circuit Judge.

Following a jury trial, Defendant was convicted of possessing more than 100 kilograms of marijuana with the intent to distribute, in violation of 21 U.S.C. § 841(a)(1) and (b)(1)(B). He was sentenced to sixty months' imprisonment followed by four years of supervised release. This appeal followed. Defendant argues the district court violated his Sixth Amendment confrontation rights when it refused to allow him to cross-examine a witness about a prior judicial determination that the witness was not credible.

BACKGROUND

Defendant was arrested on September 15, 2008, at the port of entry in Gallup, New Mexico, when a New Mexico Motor Transportation Division (MTD) inspector and police officer discovered six duffle bags containing marijuana in the trailer of the tractor-trailer Defendant was driving. Although Defendant had retired approximately four months earlier, he was driving that day as a favor to his former boss, the owner of J & J Trucking. The owner of J & J Trucking had called Defendant because one of his employees was sick and unable to work. J & J Trucking was scheduled to deliver a frozen load to the Phoenix, Arizona area, and the owner needed someone "right away" to take over this delivery. (R. Vol. 4 at 581.) Defendant agreed to cover this delivery on two conditions: first, that there would also be a return load to bring him back to Tennessee because he had a doctor's appointment he did not want to miss; and second, that he be allowed to take his girlfriend with him as he had done in the past. The owner agreed, and Defendant and his girlfriend left for Arizona immediately with the load of frozen goods. At that time, J & J Trucking already had a return load scheduled.

After Defendant left for Arizona, the return load J & J Trucking had scheduled cancelled. J & J Trucking was, however, able to locate a replacement load. On September 12, 2008, the same day Defendant delivered the frozen load in the Phoenix area, a company called Henry Company decided it needed to expedite the transportation of packaging cartons. Its plant in Kingman, Arizona, had a surplus of the cartons, while its plant in Indianapolis, Indiana, was running low. Henry Company had originally arranged for FedEx National to transport the cartons to its Indianapolis plant. But after FedEx had picked up the cartons and taken them to the FedEx facility in Phoenix, Henry Company decided it needed to find an alternative mode of transportation that would get the cartons to Indianapolis more quickly, by September 15. It decided to use a private trucking company because, unlike FedEx, a private trucking company would be able to deliver the cartons without making stops along the way. Through a somewhat lengthy chain of events, J & J Trucking was hired to transport Henry Company's load from the FedEx facility in Phoenix to the Indianapolis plant.1

Defendant arrived at the Phoenix FedEx facility on September 12 around 4:00 p.m. He backed his tractor-trailer into the dock and then went into the dispatch office to complete the required paperwork. During this time, two FedEx employees loaded the cartons onto the trailer. Only one of the two employees recalled loading the J & J Trucking trailer on September 12. He remembered he and the other employee had used forklifts to place eight pallets containing Henry Company's cartons in the front of Defendant's empty trailer. He denied loading any black bags onto the trailer. The only other person who had any involvement with the J & J Trucking tractor-trailer at the Fed Ex facility was the operations supervisor, who completed the paperwork with Defendant.2 Because the supervisor was inside the dispatch office with Defendant as the pallets were being loaded, he did not observe the loading process, although he "might have glanced" over from the office. (Id. at 189.) Neither the supervisor nor the employee smelled marijuana on the dock on September 12, 2008.

Once the cartons were loaded onto the trailer, Defendant pulled out from the dock so the trailer's swing doors, which were pinned against the dock during the loading process, could be closed. The FedEx employees testified that it would have been Defendant's responsibility to close the doors. The doors were not sealed or locked. There was conflicting testimony about whose responsibility it was to seal the trailer: the dock worker or the driver. However, it is not uncommon for trailers carrying low-value cargo, such as the cartons, to remain unlocked or unsealed.

After leaving the FedEx facility, Defendant drove to a nearby truck stop to have one of the trailer lights fixed. Unfortunately, the truck stop was too busy to fix the light anytime soon. Defendant then drove to a Petro truck stop south of Phoenix in Eloy, Arizona. The Petro was a desirable choice for truckers such as Defendant: he had a Petro passport, which enabled him to earn points and coupons that could be used for showers and food, and the Petro was more accessible than other stops in Phoenix, where it was more difficult to maneuver. However, the Petro, too, was unable to fix the light that evening. Defendant and his girlfriend spent the night, and the Petro garage fixed the light the following day.

Defendant and his girlfriend then decided to take the rest of the weekend off; Defendant was tired from the driving he had done from Memphis to Phoenix, and his girlfriend, who was suffering from cancer, was tired because of her illness. Defendant unhooked the trailer and left it behind at the Petro. He and his girlfriend then drove through the area with only the tractor, taking in the scenery. They searched for a truck stop that sold beer and wine, which the Petro did not. This led them to the Triple T in Tucson, where they spent the night eating, drinking, and watching movies. The following morning, they drove north to Holbrook, Arizona where they stayed over night.

On September 15, Defendant continued toward Memphis, Tennessee, where the owner of J & J Trucking would take over the load and deliver it to the Indianapolis plant. He was stopped at the port of entry in Gallup, New Mexico, located approximately twelve miles from the Arizona-New Mexico border. At the port of entry, an MTD inspector examined Defendant's paperwork and logbook. The inspector noticed there were some violations in Defendant's logbook, including Defendant's failure to record his trip to the Petro and the time spent off duty for repairs. After reviewing the logbook, the inspector and an MTD police officer went with Defendant to inspect the tractor-trailer. They found nothing out of the ordinary inside the cab. They then asked Defendant to open the trailer doors, which he did. Inside the trailer, the inspector and the officer discovered six black, soft-sided bags in the middle of the eight pallets of Henry Company's cartons. The inspector asked Defendant if the bags were part of the load. Defendant said, "[a]s far as I know" and "I guess they are." (Id. at 347, 372.) He explained he had not been in the trailer since before the cargo was loaded by FedEx and he had not watched what was loaded in the trailer.

The officer opened one of the bags and discovered it was filled with shrink-wrapped packages of what appeared to be marijuana. Defendant asked the officer what he had found, and the officer responded it was marijuana. Defendant reacted, stating, "[Expletive], no, I don't know nothing," "[s]omebody is [expletive] around, for sure, [expletive] around." (Id. at 372-73.) He denied any knowledge of the marijuana. Defendant was arrested and ultimately charged with possession of more than 100 kilograms of marijuana with the intent to distribute.

Before Defendant's trial began, the government filed a motion in limine to prohibit Defendant from offering evidence concerning a prior determination made by a different federal district court judge that the MTD inspector was not credible. In United States v. Variste, No. CR 06-1349 BB (D.N.M.), the district court issued a suppression order containing a finding that the court did not believe the inspector's testimony. Specifically, the Variste court found:

(R. Vol. 2 at 7.) The inspector had testified during the suppression hearing that during the inspection of the defendant's trailer at the Gallup port of entry he smelled raw marijuana. This testimony was offered to support the reasonable suspicion necessary for a search of the defendant's truck that occurred after he was subsequently pulled over by a county sheriff's deputy. The deputy had been asked by an officer at the port of entry to find his own reason to search the defendant's trailer because the MTD was unable to hold the defendant at the port of entry any longer.

In the present case, the government argued the credibility determination should be excluded under Rule 403 of the Federal Rules of Evidence because it was likely to result in a trial-within-a-trial and to confuse the issues. Defendant opposed the motion in limine, arguing that precluding him from cross-examining the inspector on the credibility determination would deprive him of his Sixth Amendment confrontation rights. In a sealed order, the district court granted the government's motion in limine. It concluded that questioning regarding the Variste court's credibility determination would likely confuse the jury, create a trial-within-a-trial, and unfairly prejudice the government. Accordingly, Defendant was prohibited from cross-examining the inspector about the credibility determination.

During trial, the government offered the inspector's testimony that when Defendant opened the doors to the trailer, the inspector "smell[ed] the odor of raw marijuana" and that the odor was "[v]ery strong." (R. Vol. 4 at 346.) The officer likewise testified that he smelled the strong odor of marijuana when the doors were opened. However, Defendant effectively impeached various aspects of the officer's testimony. First, on cross-examination, Defendant used the audio recording from the investigation of the tractor-trailer to discredit the officer's insinuation on direct examination that Defendant had inappropriately attempted to joke around with him. When pointed to specific parts of the interaction, the officer admitted his conversation with Defendant involved both of them joking around. Second, relying on the officer's written report of the incident, Defendant challenged his testimony that Defendant had acted nervous when they opened the trailer. The officer discussed the importance of being thorough in written reports and admitted he had nonetheless failed to include in his report that Defendant acted nervous. No other witnesses testified about the odor of the marijuana in the trailer.

The government relied on the strong odor of the marijuana to establish the first element of the crime — that Defendant knowingly or intentionally possessed the marijuana. Its theory of the case was that Defendant was the only one who could possibly, or at least most plausibly, have loaded the marijuana into the trailer, which demonstrated he knowingly possessed the marijuana. The government set forth this theory from the beginning of the trial, relying on the "strong odor of marijuana" in its short opening statement. (Id. at 92.) It then questioned the FedEx employees involved in loading the J & J Trucking trailer whether they noticed the smell of marijuana on the FedEx loading dock. And in its closing, the government relied extensively on the strong odor to establish the mens rea requirement, making eight separate references to the odor.

After deliberating for eight hours and having received an Allen charge, the jury returned a verdict of guilty. Defendant was sentenced to sixty months' imprisonment followed by four years of supervised release. He now appeals, arguing the district court violated his Sixth Amendment confrontation rights by prohibiting him from cross-examining the inspector on the Variste court's credibility determination.

DISCUSSION

We review de novo Defendant's claim that his Sixth Amendment confrontation rights were violated by the district court's cross-examination restriction. United States v. Robinson, 583 F.3d 1265, 1274 (10th Cir.2009).

As an initial matter, we must address the government's argument that the district court's ruling should be reviewed for abuse of discretion, not de novo. The government agrees that Sixth Amendment claims based on cross-examination restrictions are reviewed de novo. However, it urges us to reject Defendant's characterization of his claim as a constitutional one, arguing Defendant's challenge in reality involves nothing more than a "routine evidentiary ruling." (Appellee's Answer Br. at 18.) In support of this position, the government relies exclusively on United States v. Clifton, 406 F.3d 1173 (10th Cir. 2005). In Clifton, the defendant argued on appeal that the district court violated her Fifth Amendment right to due process by improperly allowing the government to introduce inadmissible evidence under the guise of impeachment. We rejected the defendant's "argument that the district court's evidentiary rulings somehow violated the Due Process Clause" and noted that we "have consistently reviewed impeachment issues, such as the one [the defendant raised], for an abuse of discretion and not under the de novo standard reserved for, among other things, constitutional questions." Id. at 1179 n. 3.

Here, unlike in Clifton, Defendant has appropriately framed his claim as a constitutional one. The crux of his argument is that the district court violated his Sixth Amendment confrontation rights when it granted the government's motion in limine, thereby precluding him from cross-examining the inspector on what he claims was an otherwise permissible subject of cross-examination — the credibility determination. This was precisely the argument Defendant made to the district court in opposition to the government's motion in limine. We are not persuaded by the government's argument that, because the district court's decision was based, in part, on what it perceived to be a danger of unfair prejudice and confusion of the issues, Defendant's claim involves only a routine evidentiary ruling. If we were to accept this reasoning, it would be possible to argue that nearly every Sixth Amendment confrontation challenge should be reviewed for abuse of discretion by pointing to the evidentiary rulings that necessarily underlie the district court's decision to preclude an area of cross-examination. We decline to open the door to such arguments. Instead, we follow our well-established precedent and review Defendant's Sixth Amendment confrontation claim de novo.

The Sixth Amendment guarantees the right of a defendant to "be confronted with the witnesses against him." U.S. Const. amend. VI. One of the primary interests secured by the Sixth Amendment's confrontation clause is the right of cross-examination. Davis v. Alaska, 415 U.S. 308, 315, 94 S.Ct. 1105, 39 L.Ed.2d 347 (1974). This is the "principal means by which the believability of a witness and the truth of his testimony are tested." Id. at 316, 94 S.Ct. 1105. A violation of this constitutional right occurs when "the defendant is prohibited from engaging in otherwise appropriate cross-examination that, as a result, precludes him from eliciting information from which jurors could draw vital inferences in his favor." United States v. Montelongo, 420 F.3d 1169, 1175 (10th Cir.2005) (internal quotation marks omitted). Stated differently, "`a defendant's right to confrontation may be violated if the trial court precludes an entire relevant area of cross-examination.'" Id. (quoting Parker v. Scott, 394 F.3d 1302, 1316 (10th Cir.2005)).

Before we can answer whether Defendant's right to confrontation was violated, we must first determine whether his proposed cross-examination of the inspector was "otherwise appropriate." Defendant argues cross-examination on the credibility determination would have been appropriate under Rule 608(b), under which the court may allow cross-examination on specific instances of a witness's conduct "if they are probative of the character for truthfulness or untruthfulness." Fed. R.Evid. 608(b). The government does not directly argue otherwise. Rather, it attempts to draw a distinction between a finding of perjury, which we presume the government believes would be admissible under Rule 608(b), and the Variste court's finding that it "[did] not believe" the inspector's testimony, which we presume the government believes would not be admissible under Rule 608(b).3 We do not find this distinction persuasive. As the Second Circuit adeptly noted, "A finding that a witness is not credible is not fundamentally different from a finding that the witness lied. It often just reflects a fact finder's desire to use more gentle language." United States v. White, 692 F.3d 235, 249 (2d Cir.2012).

Although we have not addressed the issue of whether past judicial credibility determinations are admissible under Rule 608(b), several of our sister circuits have done so and held that they are. United States v. Cedeño, 644 F.3d 79, 82-83 (2d Cir.), cert denied, ___ U.S. ___, 132 S.Ct. 325, 181 L.Ed.2d 201 (2011); United States v. Dawson, 434 F.3d 956, 957-59 (7th Cir.2006) ("[T]he decision whether to allow a witness to be cross-examined about a judicial determination finding him not to be credible is confided to the discretion of the trial judge; it is not barred by Rule 608(b), which, to repeat, is a rule about presenting extrinsic evidence, not about asking questions."); United States v. Whitmore, 359 F.3d 609, 619-22 (D.C.Cir.2004) (holding district court erred in refusing to allow the defendant to cross-examine an officer about a judge's conclusion that "I think [the officer] lied"). We find the test set forth by the Second Circuit to be particularly helpful in determining the relevancy and probative value of a prior court's finding that a witness had lied. In Cedeño, the Second Circuit set forth a list of factors a district court should consider when making this determination: "(1) whether the prior judicial finding addressed the witness's veracity in that specific case or generally; ... (2) whether the two sets of testimony involved similar subject matter"; (3) "whether the lie was under oath in a judicial proceeding or was made in a less formal context"; (4) "whether the lie was about a matter that was significant"; (5) "how much time had elapsed since the lie was told and whether there had been any intervening credibility determination regarding the witness"; (6) "the apparent motive for the lie and whether a similar motive existed in the current proceeding"; and (7) "whether the witness offered an explanation for the lie and, if so, whether the explanation was plausible." 644 F.3d at 82, 83.

In applying the Cedeño factors to the present case, we conclude the Variste court's credibility determination is relevant and highly probative. Although the Variste court addressed the inspector's credibility in that specific case only and did not find the inspector lacked credibility more generally, each of the remaining factors support our conclusion. First, the inspector's testimony in the two cases involved nearly identical subject matter. In each, he testified that upon opening the door to the defendant's tractor-trailer while performing an inspection at the Gallup port of entry, he smelled a strong odor of raw marijuana. Second, the inspector's testimony in Variste was given under oath during a suppression hearing. Third, his prior testimony was significant; it involved the central issue of the suppression hearing — whether the deputy had the requisite reasonable suspicion to perform the search of the defendant's tractor-trailer. Fourth, the testimony in Variste was relatively recent, given approximately three years before the inspector testified in the present case. And the government has not pointed to any intervening finding that the inspector was credible. Fifth, the motive to lie in both cases was the same. In each, the strong odor of marijuana was offered to support a critical determination necessary to obtain a conviction: in Variste, reasonable suspicion justifying the search, and here, whether Defendant knowingly possessed the marijuana. Sixth, and finally, the government has not offered any explanation for the inspector's implausible testimony, other than that offered to and rejected by the Variste court. Each of these factors convince us that cross-examination on the credibility determination would have been appropriate under Rule 608(b).

The government argues that even if the cross-examination was appropriate, the district court properly excluded it under Rule 403, because the questioning would likely lead to juror confusion and a trial-within-a-trial. In support of this argument, the government poses six non-exhaustive questions the jury might be left wondering had Defendant been permitted to cross-examine the inspector.4 We are not persuaded these possible questions establish a danger of confusion that substantially outweighs the probative value of the cross-examination, which, as discussed above, we conclude is very strong. They are the same type of questions that many jurors may be left wondering, particularly when witnesses are impeached during cross examination. And the government has not provided any explanation of why Defendant's proposed cross-examination was unique, setting it apart from typical impeachment on cross-examination and requiring exclusion under Rule 403. Indeed, during oral argument, the government was able to effectively address many of its own concerns through the rehabilitative testimony the inspector would have given had he been asked about the credibility determination: the government explained the inspector's testimony would have been to the effect of, "I deliberately didn't tell [the deputy] about the smell of marijuana because I wanted him to make a probable cause determination on his own." (Oral Argument at 18:47-19:05.)

As the D.C. Circuit concluded in a similar case, "[t]he district court here could have adequately guarded against any risk of unfair prejudice or undue delay by limiting cross-examination, by giving limiting instructions to the jury and by setting reasonable parameters on the government's rehabilitation of [the inspector]." Whitmore, 359 F.3d at 621 (internal citations omitted). We therefore conclude that Defendant's proposed cross-examination should not have been excluded under Rule 403. This conclusion is consistent with our precedent which "favors admission of all relevant evidence not otherwise proscribed" and recognizes "exclusion under [Rule 403] is `an extraordinary remedy [that] should be used sparingly.'" United States v. Irving, 665 F.3d 1184, 1213 (10th Cir.2011) (quoting United States v. Rodriguez, 192 F.3d 946, 949 (10th Cir.1999)) (emphasis and alterations in original).

Having concluded the proposed cross-examination was "otherwise appropriate,"5 we now turn to whether the district court's restriction violated Defendant's confrontation rights. The government argues the restriction did not because the inspector was not the government's star witness; his testimony was unrelated to Defendant's defense that he was unaware he was hauling the drugs; and the district court did not otherwise limit the ability to cross-examine the inspector. Each of these arguments is based on an incorrect application of the law.

In determining whether the Confrontation Clause has been violated, "the focus of the prejudice inquiry ... must be on the particular witness, not on the outcome of the entire trial." Delaware v. Van Arsdall, 475 U.S. 673, 680, 106 S.Ct. 1431, 89 L.Ed.2d 674 (1986). A violation occurs when a "reasonable jury might have received a significantly different impression of [the witness's] credibility had [the defendant] been permitted to pursue his proposed line of cross-examination." Id. Contrary to the government's assertion, whether the inspector was the government's "star witness" is not determinative.6 Nor does the question of whether Defendant's confrontation rights were violated hinge on whether the cross-examination was central to Defendant's defense and, therefore, would have affected the jury's verdict. "It would be a contradiction in terms to conclude that a defendant denied any opportunity to cross-examine the witnesses against him nonetheless had been afforded his right to `confront[ation]' because use of that right would not have affected the jury's verdict." Id. (alteration in original).

Here, had Defendant been permitted to cross-examine the inspector about the credibility determination, a reasonable jury might have had a significantly different impression of the inspector's credibility; the jury could have reasonably concluded the inspector was willing to exaggerate, or even fabricate, the existence of a strong odor of marijuana when necessary to support a conviction. We therefore conclude the district court's order precluding Defendant from cross-examining the inspector on the credibility determination violated his Sixth Amendment confrontation rights. This is true even though Defendant retained the ability to cross-examine the witness on other subjects. See Robinson, 583 F.3d at 1269, 1276 (10th Cir. 2009) (holding defendant's confrontation rights were violated by cross-examination restrictions even though he had the opportunity to "attempt[] to impeach [the witness] by eliciting testimony on his criminal history, the payments he received from [federal agents], and the [federal agents'] intervention on his behalf following `scrape[s]' with the law"). The government cites United States v. Rosario Fuentez, 231 F.3d 700 (10th Cir. 2000) to support its contention that the lack of other restrictions on Defendant's ability to cross-examine the inspector demonstrates Defendant's confrontation rights were not violated. But our holding in Rosario Fuentez was based on the fact the defendant "was not precluded from inquiring into an entire area of cross-examination, in this case, [the witness's] credibility. The district court allowed [the defendant] to cross-examine the officer on [other] subjects relating to his credibility." Id. at 704 (emphasis added). Here, there was no such opportunity. Even though the district court did not preclude cross-examination on other subjects related to the inspector's credibility, there were none. Unlike the defendant in Rosario Fuentez, Defendant was effectively precluded from inquiring into the entire area of the inspector's credibility.

Because we conclude Defendant's Sixth Amendment confrontation rights were violated, we must reverse Defendant's conviction unless the government can prove "that the constitutional error was harmless beyond a reasonable doubt." Van Arsdall, 475 U.S. at 681, 106 S.Ct. 1431. We have described the government's burden as an "extraordinary" one. Robinson, 583 F.3d at 1274. In deciding whether the government has met this burden, we assume the damaging potential of the cross-examination was fully realized. Van Arsdall, 475 U.S. at 684, 106 S.Ct. 1431. We then consider various factors, including "the importance of the witness' [s] testimony in the prosecution's case, whether the testimony was cumulative, the presence or absence of evidence corroborating or contradicting the testimony of the witness on material points, the extent of cross-examination otherwise permitted, and, of course, the overall strength of the prosecution's case" to determine whether the error was harmless beyond a reasonable doubt. Id. If there is a "reasonable probability" the jury would have reached a different verdict, "it necessarily follows that the district court's prohibition on cross-examination... cannot clear the high hurdle of harmlessness beyond a reasonable doubt." Robinson, 583 F.3d at 1276.

The government attempts to establish that any error was harmless by arguing the inspector's testimony was not critical to its case. It claims the only genuine issue was whether Defendant knew the marijuana was in the trailer and argues the inspector's testimony "did not bear on that question in any way." (Appellee's Answer Br. at 15.) In light of the record, the government's position is untenable. The government highlighted the "strong odor of marijuana" in its short opening statement. (R. Vol. 4 at 92.) It then questioned the FedEx employees about whether they smelled marijuana on the dock on September 12, 2008, when Defendant's tractor-trailer was loaded. And in its closing, the government relied extensively on the strong odor to satisfy the required element that Defendant knowingly possessed the marijuana, which it admitted to the jury was "the issue in this case." (Id. at 665.) The government urged the jury to "start from the premise of the one thing that gives this whole thing away ... that this load smelled strongly." (Id. at 670-71.) It explained that if the "marijuana had been on the cargo dock anywhere in that vicinity at the Phoenix location, it would have been smelly and it would have been detected." (Id. at 671.) However, neither of the FedEx employees smelled marijuana on the dock. The government argued this circumstantial evidence supported the conclusion Defendant had loaded the marijuana on the trailer in Tucson:

You'll be able to do it in secret. And that's what happened in Tucson. (Id. at 673.) The government concluded "it is clear that all the evidence only points to one person" — Defendant. (Id. at 681.)

Considering the Van Arsdall factors, we conclude there is at least a "reasonable probability" the jury would have reached a different conclusion had the jury not believed the inspector smelled the strong smell of marijuana when Defendant opened the trailer doors, that is, if the impeachment value of the cross-examination had been fully realized. First, as evidenced by the government's emphasis on the strong odor of the marijuana, the inspector's testimony was important to the government's theory that Defendant was involved in loading the marijuana on the trailer and, therefore, knowingly possessed the marijuana. Second, the testimony was not entirely cumulative. Although the officer also testified about the odor, he was effectively impeached on cross-examination. As a result, the jury could have had a more difficult time believing the marijuana had such a strong odor if it considered only the officer's testimony. See Slovik v. Yates, 556 F.3d 747, 755-56 (9th Cir.2009) (holding the unconstitutional restriction of cross-examination was not harmless, even though another witness testified similarly, because "a reasonable jury would have had a much more difficult decision ... based solely on [the one witness's] testimony"). Third, there was little other evidence supporting the government's theory that Defendant knowingly possessed the marijuana. This evidence was entirely circumstantial and consisted primarily of the presence of the marijuana in the trailer, that the FedEx employees did not see the black bags on the loading dock, and that the Defendant spent a day in Tucson, a "source city." Fourth, although Defendant was not precluded from cross-examining the inspector on other matters, he was unable to engage in any cross-examination on the inspector's credibility. Finally, the government's case was not very strong. There was no direct evidence Defendant knew the marijuana was present in the trailer. The jury was required to make inferences first about how the marijuana ended up in the trailer and then whether Defendant played a part in or was aware of that process. And we know the jury struggled in reaching its verdict. The jury deliberated for eight hours and reached a verdict only after receiving an Allen charge, further convincing us that the error was not harmless beyond a reasonable doubt. See Medina v. Barnes, 71 F.3d 363, 369 (10th Cir.1995) (considering the jury's struggle and the length of deliberation in determining whether counsel's deficient performance was prejudicial); United States v. Jean-Baptiste, 166 F.3d 102, 109 (2d Cir.1999) ("The fact that a jury initially was deadlocked and reached a verdict only after receiving an Allen charge may support an inference that the case was close."). Each of the Van Arsdall factors convince us there is a reasonable probability the jury would have reached a different result had Defendant been allowed to cross-examine the inspector on the credibility determination. At the very least, the government has not met its extraordinary burden of proving the error was harmless beyond a reasonable doubt.

CONCLUSION

For the foregoing reasons, we REVERSE Defendant's conviction and REMAND for further proceedings consistent with this opinion. Defendant's unopposed motion to seal the briefs is GRANTED.