Full opinion text
ED CARNES, Chief Judge: Because it is a document designed to govern imperfect people, the Constitution does not demand perfect trials and errors do not necessarily require the reversal of a conviction. More than thirty years ago, the Supreme Court reminded us: “As we have stressed on more than one occasion, the Constitution entitles a criminal defendant to a fair trial, not a perfect one.” Delaware v. Van Arsdall, 475 U.S. 673, 681, 106 S.Ct. 1431, 1436, 89 L.Ed.2d 674 (1986). Alexander Roy, who was convicted in federal court of five sex-related crimes involving minors, received a fair trial although not a perfect one. The error in Roy’s trial occurred when his counsel returned a few minutes late from a lunch break on the third day of the six-day trial. He missed only a small part of the testimony of the 12th of 13 government witnesses. Counsel was out of the courtroom for only seven minutes of a trial that lasted 1,884 minutes or 31.4 hours (not counting recesses and jury deliberations). That is less than one-half of one percent of the trial time. During his absence counsel missed only 18 answers out of a total of approximately 2,745 answers that were given by government witnesses during the trial. That is less than one percent of the total testimony against Roy. And the little testimony that counsel had missed was repeated in even more detail by the same witness after counsel returned to the courtroom. The parties agree that it was Sixth Amendment error for inculpatory testimony to be taken in the absence of defense counsel. Their primary disagreement is about whether it was a type of structural error for which prejudice is presumed, or trial error to which the harmless error rule applies. They also disagree about whether our review is limited to plain error and about whether the error was actually harmless. I. The Charged Crimes Roy was charged in a five-count indictment with sex crimes related to minor girls. Count 1 charged him with attempted child enticement in violation of 18 U.S.C. § 2422(b), based on his efforts to arrange a sexual encounter with someone he believed to be a 13-year-old girl in response to an internet ad posted by law enforcement. That charge did not involve any child pornography. And no questions about the Count 1 charge were asked during counsel’s brief absence. None. Counts 2-5 did involve child pornography. Each of those four counts charged Roy with knowingly possessing “any visual depiction” of child pornography in violation of 18 U.S.C. § 2252(a)(4)(B), (b)(2) (emphasis added). The difference between those four counts is based on the four different electronic devices Roy used to store his images of child pornography: his desktop computer (Count 2); his laptop computer (Count 3); his USB thumb drive (Count 4); and three of his CD-ROM discs (Count 5). All that the government had to prove under each of Counts 2-5 was that Roy knowingly possessed one or mox;e images of child pornography on the electronic device specified in that count. It could be the same image or images on each device or different images, so long as there was at least one on each device. As we will discuss in more detail in the next part, the evidence proved without dispute that there were multiple still images and video images of child pornography involving a number of different minors on each of Roy’s four electronic devices. Roy had a sexual relationship with one of the minors, and he had produced the pornographic still and video images of that child, some of which were contained on all four devices. Each of the four devices also contained other child pornography, involving different minors, that Roy had downloaded from the internet. II. The Evidence For analytical ease we break the testimony and evidence presented during the trial down into three categories: that presented before counsel’s brief absence from the trial, that presented during his absence, and that presented after he returned. A. Before Counsel’s Absence During the first two days of Roy’s six-day trial, with defense counsel present at all times, the government called 10 witnesses whose testimony focused on the attempted child enticement charge in Count 1. Their unrefuted testimony showed that Roy, a middle school teacher, set up a sexual encounter that he thought would involve a 13-year-old girl and her mother, and he drove to a pre-arranged location to meet the mother and child so that he could have sex with the child. Their testimony also showed that he went to the illicit rendezvous with condoms and a bottle of Astroglide lubricant in his pockets. Roy’s lawyer was in the courtroom for the entirety of those first two days of trial and for the presentation of all of the testimony and evidence about the crime that was charged in Count 1. He did not miss any of it on any day. On the third day of trial, before the lunch break and in counsel’s presence, there was additional testimony about Count 1, including the fact that Roy had traveled more than an hour to get to the meeting place for the purpose of having sex with a 13-year-old girl. Much of the testimony on that third morning, however, went to Counts 2-5 and concerned Roy’s sexual relationship with L.B., the girl in the pornographic images and videos that Roy himself had produced and stored, along with child pornography from the internet, on his four electronic devices specified in those four counts. That same morning, with defense counsel present, William Kulp, an agent of the Florida Department of Law Enforcement, testified without objection that L.B. was born on May 9, 1989. That means any pornography of her that was produced before May 9, 2007 is child pornography. See 18 U.S.C. § 2256(1) (defining “minor” for this purpose as anyone under 18 years of age). The principal at the high school L.B. had attended identified photos of her in various school yearbooks, three of which were admitted into evidence without objection. The principal’s testimony and those yearbook exhibits enabled the jury to compare how L.B. looked at various ages during her school years with how she looked in the pornography that Roy had produced. The third and final government witness to be called before the lunch break on the third day was Deputy Sheriff Charlie Longson, a computer forensics expert. In defense counsel’s presence, he testified extensively about his qualifications and how he examines a computer. He also testified about the user and email/messenger accounts that he had found on Roy’s desktop computer. That testimony was used, among other things, to put into evidence Roy’s email messages setting up his sexual liaison with the (fictitious) 13-year-old and the sexually oriented instant messenger conversations between Roy and (the real) L.B. that were on his computer. Longson’s testimony was interrupted by a lunch break. B. During Counsel’s Absence Defense counsel returned late from the lunch break on the third day of trial and missed seven minutes of Deputy Longson’s continuing testimony. During the seven minutes counsel was out of the courtroom, Longson gave 18 answers to the AUSA’s questions. A1 of those 18 answers concerned only six of the numerous images of child pornography, and all six of those images were of a single female subject. Those particular images of the young female were found in only one of the several file folders containing child pornography that were on Roy’s desktop computer. That folder, when discovered by Longson on Roy’s desktop computer had been labeled “2006-03-11.” On that date, L.B. indisputably would have been 16 years old. Longson testified that those six images showed “a nude white female ... bound to a table by her feet with rope” and with “an orange cloth ... secured around her neck with silver duct tape.” He also testified that the six images were taken with a Kodak v530 Zoom Digital Camera on March 10, 2005, were initially uploaded onto a computer on March 11, 2006, and were then transferred to Roy’s desktop computer on April 4, 2009. During the seven minutes while defense counsel was out of the courtroom, no exhibits were admitted into evidence and Longson did not identify L.B. as the female in the six pornographic images. C. After Counsel Returned Soon after counsel returned to the courtroom, the testimony that Longson had given during counsel’s brief absence was repeated. And it was only after counsel returned that Longson identified L.B. as the young female in the six images of pornography found on the desktop computer that he had been testifying about. This is how those events unfolded. After defense counsel entered the courtroom, the prosecutor asked the court for permission to speak with him, which the court granted. There was then a pause in the proceedings, and after the prosecutor and defense counsel had an opportunity to speak, the prosecutor approached witness Longson with 10 exhibits: the six pornographic images of L.B. that Longson had found in the “2006-03-11” folder on Roy’s desktop computer; three other pornographic images of L.B. from a different folder on that computer, which was titled “2006-12-04”; and a “contact sheet” generated by the camera showing still images from a pornographic video of L.B., also recovered from that second folder. See also infra n.3. With counsel present Longson then described in detail what each of those images depicted, and he also testified that the six images from the “2006-03-11” folder had been created on March 11, 2006, and uploaded onto Roy’s desktop computer (which Roy had acquired later) on April 4, 2009. Those six images showed the then-16-year-old L.B. “bound to a table by her feet with a ... red and white ski rope”; she was wearing an “orange hood across her head with silver duct tape secured around the neck”; there was a “dildo inserted in her vagina” and “a male’s penis ... suspended above [her] body.” During that and all the other testimony that would follow counsel was there. He was present when Longson first described the other three pornographic images of L.B. found on Roy’s desktop computer in the “2006-12-04” folder. Those images showed L.B. lying naked in a bathtub, and written in “black ink both on [her] chest between the breasts and then on [her] stomach over the nav[e]l” were the words “Alex’s Little Cunt.” (Roy’s first name, of course, is Alexander, and his roommate and L.B. both called him “Alex.”) Longson testified that those particular pornographic images were taken on December 2, 2006. On that date, L.B. indisputably would have been only 17 years old, which means she was a minor for purposes of the child pornography charges against Roy in Counts 2-5. At that point in the trial, Deputy Long-son described for the first time the contact sheet taken from the “2006-12-04” folder showing nine images from the pornographic video of L.B. A few pages later in the transcript, Longson repeated his earlier testimony that all of the images of L.B. on Roy’s desktop had been taken with a Kodak v530 Zoom Digital Camera, which is the model of camera recovered from Roy’s home during the police search. At the times all of those images of L.B. — the six in the “2006-03-11” folder and the three plus the contact sheet from the “2006-12-04” folder — were created, she was a minor for purposes of the child pornography charges against Roy in Counts 2-5 because she was under 18 years of age. See 18 U.S.C. § 2256(1). The 10 exhibits consisting of those images were admitted into evidence without objection. Being present during all of the testimony we have just recounted, Roy’s trial counsel had an opportunity to object to the testimony or to admission of the exhibits into evidence, if there were any basis for doing so. He did not object to any of it. In the presence of defense counsel, Longson also testified about finding on Roy’s desktop, laptop, thumb drives, and CD-ROM discs numerous pornographic videos of L.B. that had been made between October and December 2006 using a Kodak v530 Zoom Digital Camera. It was undisputed that L.B. would have been 17 years old, and therefore a minor, during all of that time. Some of those videos showed: L.B. bound and blindfolded with a “body net covering her body” and “a red dildo inserted into her anus”; L.B. “fully nude” with a “dildo in her vagina” while she “perform[ed] fellatio on a white male”; L.B. “fully nude” with a “vibrator in her vagina” while a white male “attempted] to have annal [sic] sex with her”; L.B. performing fellatio after removing a “schoolgirl outfit”; L.B. having sexual intercourse with a man while she was tied up; and L.B. lying “nude in [a] bathtub” with “Alex’s little cunt” scrawled across her chest and stomach while a man urinated on her. Longson described each of those videos and they were admitted into evidence. Although defense counsel was present during all of that testimony and admission of exhibits, he did not object to any of it. Deputy Longson’s testimony in defense counsel’s presence about the child pornography that he found on Roy’s desktop, laptop, USB drive, and CD-ROM discs was not limited to all of the images and videos of L.B. He also testified about finding in temporary internet files on Roy’s desktop computer several images of downloaded child pornography involving minors other than L.B., which is a subject that had not been mentioned at all during counsel’s brief absence from the courtroom. With counsel present, Longson described how one of those images of other minors showed “two or three subjects under the age of 18 engaged in sexual activity with two men.” He also described finding on Roy’s laptop a folder labeled “Girls,” which contained pornographic images of other minors and files named “kingpouge_14,” “vical6,” and “svet„16.” Longson testified that he had found five images of child pornography featuring minors other than L.B. on Roy’s USB thumb drive. And he testified that he had found on Roy’s CD-ROM discs multiple pornographic images of minors other than L.B., which were copies of images on Roy’s other devices. All of those were pornographic images of minors other than L.B., and all of them were admitted into evidence. Although he was present during all of that testimony, defense counsel did not object to any of it. Any one of those pornographic images of minors other than L.B. was enough by itself to prove the crime of possession of child pornography in violation of 18 U.S.C.' § 2252(a)(4)(B), (b)(2), which is the crime charged in Counts 2-5 of the indictment. Once the prosecution completed its direct examination of Longson, defense counsel cross-examined him over the course of 45 pages of the trial transcript. He attempted to challenge Longson’s testimony that the images and videos of L.B. were created when she was under the age. of 18. His challenge fell short, however, because Longson explained that data embedded in the images and videos of L.B. showed that they had been taken on a date when L.B. was a minor. Defense counsel did not even attempt during cross-examination or at any other time to challenge Longson’s testimony about the pornographic images involving minors other than L.B. On the fourth day of trial, the government called its last witness and then rested. The defense called a few witnesses, including Robert Deane Moody, its own computer forensics expert. He testified that there were reported problems with the battery life of the Kodak camera model that Roy had used to produce the pornographic images of L.B., which would cause the camera’s internal clock to reset to its default date and time if the camera’s battery went dead. If the internal clock in the camera used to create the images of L.B. had reset, in his opinion it was possible that the creation dates that Deputy Long-son had noted for the L.B. images and videos might be inaccurate. Moody conceded, however, that the problems he had described were not necessarily present in all Kodak v530 cameras, and he conceded that Roy’s camera might not have had any battery issues anyway. He admitted that the dates applied by a user to the computer folders in which the L.B. images were stored (i.e., “2006-03-11,” “2006-10-13,” and “2006-12-04”) were all consistent with the creation dates that the camera had automatically embedded in those images themselves. Moody also admitted that the images and videos were numbered sequentially and none of them showed any signs of having reverted back to an earlier date. III. The Facts Concerning Counsel’s Brief Absence We know only these facts about counsel’s absence. On the third day of trial during the testimony of Deputy Longson, who was the 12th of 13 government witnesses, the judge announced the lunch break: “Okay. So let’s go ahead and break for lunch and ask you to be back at 1:30.” The jury left the courtroom at 12:33. The next thing in the transcript is this parenthetical notation by the court reporter: “(Court recessed at 12:34 p.m., and proceedings continued, without the presence of defense counsel, at 1:29 p.m.).” The testimony of Deputy Longson resumed at 1:29 p.m. and continued for two- and-a-half transcript pages, consisting of 18 questions and answers, after which the following occurred: [AUSA]: Your Honor, may I have a moment while I approach Counsel? (Defense counsel entered the courtroom at 1:36 p.m.) (Pause.) [AUSA]: Thank you, Your Honor. May I approach, Your Honor? The Court: All right. [AUSA]: I’m showing the witness Government’s Exhibits 73-01 through 73-10. To recap, after lunch the trial resumed one minute earlier than it had been scheduled to, and defense counsel returned six minutes later than the time he had been instructed to be there. As a result, he missed seven minutes of a trial that lasted a total of 1,884 minutes or 31.4 hours (not counting recesses and jury deliberations), which means he was present during 99.6 percent of the trial. Counsel missed hearing only 18 answers given by one of the 13 government witnesses against him, who collectively gave a total of approximately 2,745 answers. Even if we consider only the testimony of Deputy Longson, the witness who was on the stand when he returned late, counsel missed only three of the 175 pages of Longson’s total testimony (which consisted of 111 pages of direct examination, 45 pages of counsel’s cross-examination, and 19 pages of redirect examination). We know that from the record. We do not know why counsel returned late from lunch. We also do not know if he realized when he walked in late that some testimony had been taken in his absence, either because he heard testimony being given, or he saw that there was a witness on the stand and the AUSA was up, or because his client who had been present told him what had happened. And we do not know if either the AUSA or the judge realized that defense counsel was absent when the trial resumed after lunch. One thing that we do know is that neither party wants us to take the necessary steps to find out any of those facts. Both sides insist that instead of remanding for an evidentiary hearing to determine all of the other facts about counsel’s brief absence, including who knew what and when, we should decide the appeal solely on the basis of the facts that are already in the record. We will. IV. An Assumption to Simplify the Analysis And Focus on the Harmless Error Issue The government argues that we should review only for plain error and that there isn’t any. See United States v. Duncan, 400 F.3d 1297, 1301 (11th Cir. 2005) (“We have discretion to correct an error under the plain error standard where (1) an error occurred, (2) the error was plain, (3) the error affected substantial rights, and (4) the error seriously affects the fairness, integrity or public reputation of judicial proceedings.”). Absent any knowledge of why defense counsel was absent, whether the AUSA or judge realized he was not present, about what counsel realized or didn’t when he walked in late, and about whether he took some ameliorative action not reflected in the transcript, we will not apply the plain error rule or remand for any findings necessary to decide if it is applicable. Instead, in order to simplify our analysis, we will indulge the assumption that the plain error rule does not apply even though there was no contemporaneous objection. We can indulge that assumption because even with it the result is the same. Given that scope of review, we do agree with Roy that absent evidence of an attempt to deliberately inject error into the record and without a waiver from the defendant, it is a violation of the Sixth Amendment for inculpatory testimony to be taken from a government witness without the presence of at least one of the defendant’s counsel, regardless of whether the judge or the AUSA noticed that counsel was not there. We do not, however, agree with Roy that prejudice is presumed and reversal is automatic. Instead, for the reasons that follow we hold that the harmless error rule is applicable to this brief absence of counsel from the courtroom, and that the absence was harmless beyond a reasonable doubt in this case. V. Analysis: Why the Harmless Error Rule Applies and the Rare Exceptions to It Do Not Given our assumptions in Roy’s favor, the outcome turns on whether the error in this case, like most constitutional errors, is one to which the harmless error rule applies or instead is one of those rare cases where the presumption of prejudice applies. If counsel’s brief absence is a type of structural error, we presume prejudice and there will be no room for the application of the harmless error rule. If it is not structural error, and no other rare exception requiring that prejudice be presumed fits, the harmless error rule applies. And, as we will explain later, the error was harmless beyond a reasonable doubt. See infra Part VI. A. The Importance of the Harmless Error Rule and How Pervasively It Applies The harmless error rule serves vital interests, chief of which is conserving scarce judicial resources by avoiding pointless retrials. Applying the rule to determine whether error, including constitutional error, affected the result of a trial is also essential to avoid a “sporting theory of justice” and a regime of gotcha review. See United States v. Agurs, 427 U.S. 97, 108, 96 S.Ct. 2392, 2400, 49 L.Ed.2d 342 (1976) (quotation marks omitted). “Reversal for error, regardless of its effect on the judgment, encourages litigants to abuse the judicial process and bestirs the public to ridicule it.” Van Arsdall, 475 U.S. at 681, 106 S.Ct. at 1436 (quotation marks omitted). The Supreme Court has explained that the harmless error rule “promotes public respect for the criminal process by focusing on the underlying fairness of the trial.” Neder v. United States, 527 U.S. 1, 18, 119 S.Ct. 1827, 1838, 144 L.Ed.2d 35 (1999) (quotation marks omitted); see also Johnson v. United States, 520 U.S. 461, 470, 117 S.Ct. 1544, 1550, 137 L.Ed.2d 718 (1997) (reviewing only for plain error a violation of the Sixth Amendment right to jury, trial and deciding that “there is no basis for concluding that the error seriously affected the fairness, integrity or public reputation of judicial proceedings. Indeed, it would be the reversal of a conviction such as this which would have that effect.”) (quotation marks and alterations omitted); see also 28 U.S.C. § 2111 (“On the hearing of any appeal or writ of certiorari in any case, the court shall give judgment after an examination of the record without regard to errors or defects which do not affect the substantial rights of the parties.”); Shinseki v. Sandera, 556 U.S. 396, 407-08, 129 S.Ct. 1696, 1705, 173 L.Ed.2d 532 (2009) (construing § 2111 “as expressing a congressional preference for determining ‘harmless error’ without the use of presumptions insofar as those presumptions may lead courts to find an error harmful, when, in fact, in the particular case before the court, it is not”). We are, after all, talking about “the harmless error rule,” not “the harmless error exception.” Because errorless trials are not expected, much less required, harmless error analysis is the rule, not the exception. How broadly the rule applies is evident from the Supreme Court’s observation that: “Since this Court’s landmark decision in Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), in which we adopted the general rule that a constitutional error does not automatically require reversal of a conviction, the Court has applied harmless-error analysis to a wide range of errors and has recognized that most constitutional errors can be harmless.” Arizona v. Fulminante, 499 U.S. 279, 306, 111 S.Ct. 1246, 1263, 113 L.Ed.2d 302 (1991) (emphasis added). The Court drove home that point by listing in the Fulminante opinion 16 different constitutional violations that it had held are subject to the harmless error rule. And the decision in Fulminante became a 17th example by holding that admission of a coerced confession is another error that can and should be reviewed for harmlessness. Id. at 306-09, 111 S.Ct. at 1263-64. See infra at 1167-68. The dissenting opinion seeks to sweep away the important point that the Supreme Court made in Fulminante when it listed 16 constitutional errors (plus the one in that case itself) that have been held to be subject to harmless error analysis instead of a presumption of prejudice. See infra at 1167. The dissent would replace the Supreme Court’s instruction in Fulmi-nante about the breadth of the harmless error rule with an alternative message that only “unimportant and insignificant” constitutional errors are subject to harmless error review under Chapman. See Dissenting Op. at 1230. But under Chapman constitutional errors are not to be classified by the importance or significance of the constitutional right that was violated but by the effect of the violation “in the setting of a particular case.” See Chapman, 386 U.S. at 22, 87 S.Ct. at 827. That is why, as the Court’s Fulminante list shows, the harmless error rule has been applied to all types of constitutional errors, including: defects of every sort in jury instructions; restrictions on the right to cross-examine adverse witnesses; improper comments on the right to remain silent at trial; violation of the right of the defendant to be present at trial; admission of a coerced confession; admission of evidence in violation of the right to counsel; and denial of counsel at a preliminary hearing. See infra at 1167. Those violations do not involve “unimportant and insignificant” constitutional rights, but the Court has applied the harmless error rule to them nonetheless. B. The Cronic Exception For virtually every rule of law, however, there is an exception or two, sometimes more. One of those exceptions at issue in this appeal is the Cronic exception, which provides that prejudice is to be presumed, and therefore the harmless error rule does not apply, when a criminal defendant has been completely denied the right to counsel for a critical stage of the trial, which is an error that contaminates the entire proceeding. See United States v. Cronic, 466 U.S. 648, 659 & n.25, 104 S.Ct. 2039, 2047 & n.25, 80 L.Ed.2d 657 (1984). When an error of that magnitude happens, we do not ask whether the error was harmless; we irrebutably presume that it was harmful. See id.; see also Bell v. Cone, 535 U.S. 685, 695-96, 122 S.Ct. 1843, 1850-51, 152 L.Ed.2d 914 (2002) (noting that Cronic “identified three situations implicating the right to counsel” in which prejudice to the defense could be presumed). Roy’s primary contention is that his counsel’s brief absence from the courtroom is Cronic error. It is not. The Cronic decision limited the presumption of prejudice to cases where defense counsel “entirely fails to subject the prosecution’s case to meaningful adversarial testing” in the trial or where there is “the complete denial of counsel” at a “critical stage of [the] trial.” Cronic, 466 U.S. at 659, 104 S.Ct. at 2047 (emphasis added). Roy has never contended, and could not contend, that his counsel entirely failed to subject the prosecution’s case to meaningful testing. Defense counsel was present during 99.6 percent of Roy’s trial, and he vigorously represented Roy. Among other things, he cross-examined nine of the government’s 13 witnesses, including Deputy Longson whom he cross-examined for 45 pages of the trial transcript. Counsel also called his own competing expert witness in an attempt to rebut Longson’s testimony. And he gave a vigorous closing argument. In sum, Roy’s counsel did “subject the prosecution’s case to meaningful adversarial testing.” Id. Instead of questioning the effectiveness of his representation, Roy contends his case falls within the Cronic exception because his counsel’s brief absence during one small part of the testimony of one of the 13 government witnesses against him amounts to denial of counsel during a critical stage of the trial. We turn now to the critical stage requirement that must be met before an error will be found to fit within the Cronic exception to the prejudice requirement and the harmless error rule. See id. The last time that we sat en banc in a case involving a Cronic issue, we emphasized that the exception applied “to only a very narrow spectrum of cases” where “the defendant was in effect denied any meaningful assistance at all.” Stano v. Dugger, 921 F.2d 1125, 1153 (11th Cir. 1991) (en banc) (emphasis added) (quotation marks omitted); see United States v. Kaid, 502 F.3d 43, 46 (2d Cir. 2007) (expressing “reluctance to extend a rule of per se prejudice in any new direction”) (quotation marks omitted). And we emphasized that the burden of establishing that an error warrants Cronic’s presumption of prejudice is “a very heavy one.” Stano, 921 F.2d at 1153 (quotation marks omitted). The difficulty of carrying that “very heavy” burden and the “very narrow” scope of the Cronic exception are evident from the fact that the Supreme Court has repeatedly refused to find it applicable. The Court has held that the Cronic exception did not apply, and the usual showing of actual prejudice was required, where trial counsel failed to present any mitigating evidence or make any final argument during the penalty phase of a capital trial. Bell, 535 U.S. at 692-98, 122 S.Ct. at 1849-52. And the Court has held that the Cronic exception did not apply and a showing of actual prejudice was required where trial counsel, without the defendant’s consent, conceded that the defendant was guilty of capital murder as part of his strategy to avoid a death sentence. Florida v. Nixon, 543 U.S. 175, 178, 190-92, 125 S.Ct. 551, 555, 562-63, 160 L.Ed.2d 565 (2004). Only once in the 30 years since the Cronic decision was issued has the Supreme Court applied Cronic to presume prejudice. See Penson v. Ohio, 488 U.S. 75, 88, 109 S.Ct. 346, 354, 102 L.Ed.2d 300 (1988) (holding that “the presumption of prejudice must extend as well to the denial of counsel on appeal” when the granting of an attorney’s motion to withdraw had left the petitioner “entirely without the assistance of counsel on appeal”). The scope of the Cronic exception is that narrow; the burden of showing it applies is that heavy. Even in Cronic itself the Court did not find that the Cronic exception to the harmless error rule applied. That case involved a woefully inexperienced, young attorney who had been appointed to serve as counsel less than a month before trial in a complex mail fraud case, a case that the government had investigated for over four- and-a-half years during which it had reviewed thousands of documents. Cronic, 466 U.S. at 649, 104 S.Ct. at 2041. Despite those extreme facts, the Supreme Court refused to presume prejudice, requiring instead that the defendant show that he actually was prejudiced. Id at 662-66, 104 S.Ct. at 2049-50. The Court remanded the case for the court of appeals to determine whether the defendant could establish deficient performance and prejudice, as required by Strickland v. Washington. Id. at 666-67, 104 S.Ct. at 2051. The Supreme Court’s insistence on confining the Cronic exception within narrow boundaries is evident from the fact that in Nixon, Bell, and Cronic itself the Court reversed the decisions of lower courts that had held the exception applied and had presumed prejudice. See Nixon, 543 U.S. at 189-93, 125 S.Ct. at 561-63; Bell, 535 U.S. at 688, 702, 122 S.Ct. at 1847, 1854; Cronic, 466 U.S. at 666-67, 104 S.Ct. at 2051. And in all of those cases, the risk of prejudice to the defendant was much greater than the risk of prejudice to Roy from his lawyer’s seven-minute absence during a six-day trial. One way that the Supreme Court has ensured that the Cronic exception will remain rare, the scope of the decision will be narrow, and the burden of establishing the exception will be heavy is by requiring that there be a complete denial or total failure of counsel, if not at trial generally, at least at a critical stage of the prosecution. See Cronic, 466 U.S. at 659, 104 S.Ct. at 2047 (“The presumption that counsel’s assistance is essential requires us to conclude that a trial is unfair if the accused is denied counsel at a critical stage of his trial.”); see also Bell, 535 U.S. at 697, 122 S.Ct. at 1851 (noting that counsel’s failure to test the prosecution’s case “at specific points” does not rise to the level of Cronic error). In the Cronic opinion itself, the Court’s examples of a critical stage include Hamilton v. Alabama, 368 U.S. 52, 54-55, 82 S.Ct. 157, 159, 7 L.Ed.2d 114 (1961), where prejudice was presumed when the defendant was entirely denied any counsel throughout all of his arraignment, and White v. Maryland, 373 U.S. 59, 59-60, 83 S.Ct. 1050, 1051, 10 L.Ed.2d 193 (1963), where prejudice was presumed after the defendant was entirely denied counsel throughout all of his preliminary hearing. See Cronic, 466 U.S. at 659 n.25, 104 S.Ct. at 2047 n.25; see also Strickland v. Washington, 466 U.S. 668, 692, 104 S.Ct. 2052, 2067, 80 L.Ed.2d 674 (1984) (“Actual or constructive denial of the assistance of counsel altogether is legally presumed to result in prejudice.”) (emphasis added). Roy’s position depends on his proposition that what took place during the seven minutes when his counsel was out of the courtroom is unto itself a critical stage of the trial. If the 18 answers that counsel missed hearing from one government witness, but of a total of 2,745 answers from 13 government witnesses during the trial, do not by themselves constitute a separate stage of the trial, Roy’s Cronic argument fails. So Roy argues, as he must, that what occurred during those seven minutes must be considered by itself to be “a critical stage of his trial.” Cronic, 466 U.S. at 659, 104 S.Ct. at 2047. What, then, is a “critical stage” of a trial? We, like the Sixth Circuit, “would welcome a comprehensive and final one-line definition of ‘critical stage’ ” for the purposes of determining whether error is Cronic error. Van v. Jones, 475 F.3d 292, 312 (6th Cir. 2007). None exists, as that court recognized. H. We do not, however, need a comprehensive or pithy definition of the term to conclude that the brief period during which Roy’s counsel was absent from the courtroom is not itself a critical stage of the trial. If we held that seven minutes of a six day trial, and 18 answers from one of 13 government witnesses, who gave a total of 2,745 answers during their testimony, amounts to a stage of a trial, we would have to conclude that the presentation of the government witnesses at Roy’s trial was a collection of 152 separate critical stages (2,745 ÷ 18 = 152.5) not even counting other parts of the trial. If we did that, Cronic’s “very narrow” exception would be very broad, contrary to what the Supreme Court and this Court stated. See Stano, 921 F.2d at 1153. If 18 answers from one of 13 witnesses against a defendant were enough to be a critical stage, what would not be? Would a single question and inculpatory answer from a government witness be enough to constitute a critical stage of the trial? Under Roy’s extreme view it would be. He argues that: “The presentation of inculpa-tory testimony by a government witness is a critical stage of trial.” En Banc Br. of Appellant at 14. The dissenting opinion agrees with that view. If counsel misses even one inculpatory answer from a government witness, in the dissent’s view that’s it, irreparable error has been committed no matter what happens in the rest of the trial. But it cannot be the law that every inculpatory answer given by every government witness (or defense witness on cross-examination) is a separate stage of the proceedings against the defendant. Trials don’t consist of thousands of critical stages. Although the brevity of counsel’s absence in this case and how little he missed is striking, it’s not merely the fleeting nature of the absence that convinces us that counsel was not gone during an entire “stage of [the] trial.” See Cronic, 466 U.S. at 659, 104 S.Ct. at 2047. Length alone does not always define a stage of a trial. Depending on the circumstances, an arraignment could take 10 minutes or less, although it is a critical stage. See Bell, 535 U.S. at 695-96, 122 S.Ct. at 1851. The Supreme Court has instructed us that it has used the term “critical stage” “to denote a step of a criminal proceeding, such as arraignment, that held significant consequences for the accused.” Bell, 535 U.S. at 695-96, 122 S.Ct. at 1851. And decision after decision shows that what the Court means when it does use the term “stage” for Cronic purposes is a qualitatively distinct, discrete, and separate phase or step of a criminal proceeding where the defendant has a right to counsel, such as an arraignment, a post-indictment lineup, a preliminary hearing, a plea hearing, closing arguments as a whole, or a sentence proceeding as a whole. See Montejo v. Louisiana, 556 U.S. 778, 786, 129 S.Ct. 2079, 2085, 173 L.Ed.2d 955 (2009) (describing post-indictment interrogation as a critical stage); Iowa v. Tovar, 541 U.S. 77, 87, 124 S.Ct. 1379, 1387, 158 L.Ed.2d 209 (2004) (“A plea hearing qualifies as a ‘critical stage.’ ”); Gardner v. Florida, 430 U.S. 349, 358, 97 S.Ct. 1197, 1205, 51 L.Ed.2d 393 (1977) (“[Sentencing is a critical stage of the criminal proceeding at which [the defendant] is entitled to the effective assistance of counsel.”); Gilbert v. California, 388 U.S. 263, 272, 87 S.Ct. 1951, 1956, 18 L.Ed.2d 1178 (1967) (“[A] post-indictment pretrial lineup ... is a critical stage of the criminal prosecution....”); White, 373 U.S. at 59-60, 83 S.Ct. at 1051 (“Whatever may be the normal function of the ‘preliminary hearing’ under Maryland law, it was in this case as ‘critical’ a stage as arraignment. ...”); Hamilton, 368 U.S. at 53, 82 S.Ct. at 158 (describing arraignment as “a critical stage in a criminal proceeding”). In conformity with what the Supreme Court has done in this area, our sister circuits generally treat “stage” in “critical stage” as meaning either a self-contained proceeding or a discrete and separately identifiable portion of a larger proceeding. See, e.g., United States v. Ross, 703 F.3d 856, 873-74 (6th Cir. 2012) (deciding that a competency hearing is a critical stage); McNeal v. Adams, 623 F.3d 1283, 1285, 1289 (9th Cir. 2010) (after considering several factors that might “make a proceeding a critical stage,” holding that a hearing on a motion to compel the defendant to provide a DNA sample is not a critical stage) (emphasis added); McDowell v. Kingston, 497 F.3d 757, 762-63 (7th Cir. 2007) (explaining that no Supreme Court authority indicates “that [a defendant’s] testimony, isolated from the rest of his defense, constitutes a critical stage of the litigation,” and holding that even the complete testimony of the defendant is not a critical stage); Harrington v. Gillis, 456 F.3d 118, 132 (3d Cir. 2006) (noting that “an appeal is a critical stage of criminal proceedings”) (emphasis added); United States v. Sanchez-Barreto, 93 F.3d 17, 20 (1st Cir. 1996) (noting that a “plea withdrawal hearing” is a critical stage) (emphasis added). Those decisions of the Supreme Court and of other circuits are consistent with the everyday definition of “stage” as “a single step or degree in a process; a particular phase, period, position, etc., in a process, development, or series.” Stage, Random House Webster’s Unabridged Dictionary (2d ed. 2001) 1853-54. In our lives, as well as throughout the law, when we refer to “stages” we do not mean fleeting moments or small parts of events. Instead, we use the word to refer to larger, discrete component parts of a process that share a common characteristic. For example, adolescence is a stage of life, but we would never speak or think of every minute, hour, or day during adolescence by itself as a separate or discrete stage of life. The 18 questions and answers that Roy’s counsel missed do not fit any accepted definition of “stage” or “critical stage.” They do not constitute a separate step in the process of the trial, or a discrete phase of it. Not only are they not a stage of the trial, those 18 questions and answers are not even an identifiable stage of Deputy Longson’s testimony. They are just a small part of it — only three transcript pages out of 177 total pages of his testimony. Nothing but counsel’s absence marks the 18 questions to Longson as different from all of the others put to him before lunch or all of those put to him after he returned to the courtroom following lunch. They are all questions and answers of the same type as those that preceded and followed them, and they occurred during direct examination of the same one of the 13 government witnesses, asked by the same government lawyer. The 18 questions and answers counsel missed are just a small part of the more than 2,500 that occurred during the six-day trial. Not only that, but all of those 18 questions were repeated after counsel returned to the courtroom. The only defining characteristic of what took place in the trial during the seven minutes while Roy’s counsel was absent is that it occurred while Roy’s counsel was absent. Roy would have us define “stage” to equate with the absence of an attorney so that anything that happened in a trial during the absence of an attorney, however brief it was, would be a stage of the trial. That definition is hopelessly circular. Because the brief period during which Roy’s counsel was absent is not itself a “stage of his trial,” Roy did not suffer “the complete denial of counsel” for “a critical stage of his trial.” Cronic, 466 U.S. at 659, 104 S. Ct at 2047. For that reason, there was no Cronic error in this ease. We will discuss the Cronic “critical stage” arguments of Roy and the dissent now. After doing that, we will turn to the related but different question of whether a presumption of prejudice should arise when defense counsel is absent from a substantial portion of the trial. 1. The Geders, Herring, and Brooks Decisions The dissenting opinion relies on Geders v. United States, 425 U.S. 80, 96 S.Ct. 1330, 47 L.Ed.2d 592 (1976), Herring v. New York, 422 U.S. 853, 95 S.Ct. 2550, 45 L.Ed.2d 593 (1975), and Brooks v. Tennessee, 406 U.S. 605, 92 S.Ct. 1891, 32 L.Ed.2d 358 (1972), which it contends involved “the denial of counsel ‘at a critical stage of ... trial.’ ” Dissenting Op. at 1244 & n.14 (alteration in original) (emphasis omitted) (quoting Cronic, 466 U.S. at 659, 104 S.Ct. at 2047). That interpretation ignores the unique type of constitutional violations those cases involved and it ignores what the Court later said about those decisions. See Perry v. Leeke, 488 U.S. 272, 279-80, 109 S.Ct. 594, 599-600, 102 L.Ed.2d 624 (1989). In its pre-Cronic decision of Geders, the Court applied a presumption of prejudice to a Sixth Amendment violation that occurred when the trial court barred defense counsel from advising or otherwise assisting his client during a 17-hour recess. 425 U.S. at 91, 96 S.Ct. at 1337. The order had prevented the defendant from discussing important matters with counsel, including “tactical decisions to be made and strategies to be reviewed.” Id. at 88, 96 S.Ct. at 1335. The Geders decision did not explicitly apply the “critical stage” rule or analysis; in fact, the opinion does not mention the term “critical stage” or even the word “stage.” Instead, as the Court explained later, Geders was one of a line of decisions presuming prejudice where a defense attorney was prevented from, or impeded in, rendering assistance of counsel to his client because of an unconstitutional statute or court order. See Perry, 488 U.S. at 279-80, 109 S.Ct. at 599-600. Recognizing that special subtype of Sixth Amendment violation, as the Court pointed out in Perry, is consistent with what Strickland itself held. Id. at 279, 109 S.Ct. at 599. While shortcomings and failures of counsel require a petitioner to show prejudice from the deficient performance, “direct governmental interference with the right to counsel is a different matter.” Id. The Perry Court quoted the following passage from Strickland to drive home the point: Government violates the right to effective assistance when it interferes in certain ways with the ability of counsel to make independent decisions about how to conduct the defense. See, e.g., Geders v. United States, 425 U.S. 80, 96 S.Ct. 1330 [47 L.Ed.2d 592] (1976) (bar on attorney-client consultation during overnight recess); Herring v. New York, 422 U.S. 853, 95 S.Ct. 2550 [45 L.Ed.2d 593] (1975) (bar on summation at bench trial); Brooks v. Tennessee, 406 U.S. 605, 612-13, 92 S.Ct. 1891, 1895 [32 L.Ed.2d 358] (1972) (requirement that defendant be first defense witness); Ferguson v. Georgia, 365 U.S. 570, 593-96, 81 S.Ct. 756, 768-70 [5 L.Ed.2d 783] (1961) (bar on direct examination of defendant). Id. at 280, 109 S. Ct. at 599 (citations altered) (quotation marks omitted). The statutory or court-ordered interference exception to the prejudice requirement that was applied in Geders, Herring, and Brooks, that was recognized in Strickland, and that was discussed in Perry, does not apply in this case and does not govern our critical stage analysis. No statute or court-ordered bar kept Roy’s trial counsel out of the courtroom for those seven minutes following lunch on the second day of trial. And no statute or court order interfered with the ability of Roy’s counsel to make independent decisions about how to conduct the defense. 2. The Gonzalez-Lopez, Woods, and Williams Decisions The dissenting opinion also relies heavily on the Supreme Court’s decision in United States v. Gonzalez-Lopez, 548 U.S. 140, 126 S.Ct. 2557, 165 L.Ed.2d 409 (2006), which did not involve an attorney’s brief absence from the courtroom. Instead, it involved a complete violation of “the right of a defendant who does not require appointed counsel to choose who will represent him.” Id. at 144, 126 S.Ct. at 2561; see id. at 143-44, 126 S.Ct. at 2561 (holding that the district court’s erroneous rulings “violated respondent’s Sixth Amendment right to paid counsel of his choosing”); ⅛ at 146, 126 S.Ct. at 2562 (“[T]he right at stake here is the right to counsel of ehoice[.]”); id. at 147, 126 S.Ct. at 2563 (“The right to select counsel of one’s choice, by contrast [to the right to effective assistance of counsel], has never been derived from the Sixth Amendment’s purpose of ensuring a fair trial.”); id. at 152, 126 S.Ct. at 2566 (“[T]he Government has conceded that the District Court here erred when it denied respondent his choice of counsel.”). The deprivation of the right to retained counsel of choice in Gonzalez-Lopez was anything but momentary; it lasted longer than the trial itself. It was complete, lasting throughout the entirety of the opening statements, the presentation of all of the prosecution’s case, the presentation of all of the defense case, the closing arguments, the jury instructions, the return of the verdict, and the post-verdict proceedings. Id. at 142-14, 126 S.Ct.-at 2560-61. As the Supreme Court noted, “the deprivation of choice of counsel pervade[d] the entire trial.” Id. at 150, 126 S.Ct. at 2565. As a result, the start-to-finish “erroneous deprivation of the right to counsel of choice” in Gonzalez-Lopez had “consequences that are necessarily unquantifiable and indeterminate” and “unquestionably qualifies as ‘structural error.’ ” Id. at 150, 126 S Ct. at 2564 (quotation marks omitted). The Supreme Court explained in some detail why it would be impossible to apply the harmless error rule and gauge the prejudicial effect of depriving a defendant of the attorney he had retained and forcing him to use a different one during the entire trial and post-trial stages: Different attorneys will pursue different strategies with regard to investigation and discovery, development of the theory of defense, selection of the jury, presentation of the witnesses, and style of witness examination and jury argument. And the choice of attorney will affect whether and on what terms the defendant cooperates with the prosecution, plea bargains, or decides instead to go to trial. In light of these myriad aspects of representation, the erroneous denial of counsel bears directly on the “framework within which the trial proceeds,” Fulminante, supra, at 310, 111 S.Ct. 1246 — or indeed on whether it proceeds at all. It is impossible to know what different choices the rejected counsel would have made, and then to quantify the impact of those different choices on the outcome of the proceedings. Many counseled decisions, including those involving plea bargains and cooperation with the government, do not even concern the conduct of the trial at all. Harmless-error analysis in such a context would be a speculative inquiry into what might have occurred in an alternate universe. Id. at 150, 126 S. Ct. at 2564-65. The Court also explained the difference between the denial of retained counsel of choice and more typical ineffective assistance violations: [I]f and when counsel’s ineffectiveness “pervades” a trial, it does so (to the extent we can detect it) through identifiable mistakes. We can assess how those mistakes affected the outcome. To determine the effect of wrongful denial of choice of counsel, however, we would not be looking for mistakes committed by the actual counsel, but for differences in the defense that would have been made by the rejected counsel — in matters ranging from questions asked on voir dire and cross-examination to such intangibles as argument style and relationship with the prosecutors. We would have to speculate upon what matters the rejected counsel would have handled differently — or indeed, would have handled the same but with the benefit of a more jury-pleasing courtroom style or a longstanding relationship of trust with the prosecutors. And then we would have to speculate upon what effect those different choices or different intangibles might have had. The difficulties of conducting the two assessments of prejudice are not remotely comparable. Id. at 150-51, 126 S. Ct. at 2565. Those explanations underscore how distinguishable the Gonzalez-Lopez case is from this one. None of the Supreme Court’s reasoning about why it is impossible to gauge the prejudicial impact of forcing a different attorney on the defendant throughout the entire trial and post-trial stages of a case applies to a seven-minute absence of counsel during a six-day trial when the missed testimony was not only transcribed for review but was also repeated in the presence of counsel after he returned (and as repeated was transcribed again). The momentary absence of counsel from the courtroom in this case is entirely different from the complete denial of counsel of choice throughout the Gonzalez-Lopez case. A momentary absence, unlike a complete denial of counsel of choice, does not affect the choice of “strategies with regard to investigation and discovery, development of the theory of defense, selection of the jury, presentation of the witnesses, and style of witness examination and jury argument.” Id. at 150, 126 S.Ct. at 2564. It does not “affect whether and on what terms the defendant cooperates with the prosecution, plea bargains, or decides instead to go to trial.” Id. It does not require us to consider, as courts would with a complete denial of counsel of choice, “such intangibles as argument style and relationship with the prosecutors,” or what things the denied counsel “would have handled differently — or indeed, would have handled the same but with the benefit of a more jury-pleasing courtroom style or a longstanding relationship of trust with the prosecutors.” Id. at 151, 126 S.Ct. at 2565. The denial of counsel of choice “bears directly on the framework within which the trial proceeds — or indeed on whether it proceeds at all.” Id. at 150, 126 S.Ct. at 2564-65 (quotation marks and citation omitted). The momentary absence of Roy’s counsel from the courtroom does not. To borrow the Supreme Court’s words, “[t]he difficulties of conducting the two assessments of prejudice are not remotely comparable.” Id. at 151, 126 S.Ct. at 2565. They are not comparable because what Roy’s momentarily absent counsel would have done, or should have done, had he been present are “identifiable mistakes,” and “[w]e can assess how those mistakes affected the outcome.” Id. at 150-51, 126 S.Ct. at 2565; see infra Part VI (explaining why the error was harmless beyond a reasonable doubt in this ease). The dissent also goes astray in its reading of Woods v. Donald, 575 U.S. -, 135 S.Ct. 1372, 191 L.Ed.2d 464 (2015) (per curiam), a decision that actually reversed a grant of habeas relief based on a lower court’s holding that Cronic error occurred when defense counsel was absent for 10 minutes during the testimony of a prosecution witness. See Dissenting Op. at 1240. The Sixth Circuit had held that the state court decision denying the petitioner habe-as relief because of that 10-minute absence was contrary to, or involved an unreasonable application of, clearly established federal law as determined by the Supreme Court within the meaning of 28 U.S.C. § 2254(d)(1). 135 S.Ct. at 1375. In reversing the Sixth Circuit, the Supreme Court pointed out that none of its own holdings have addressed defense counsel’s absence during the presentation of testimony that is irrelevant to the defendant’s theory of the case. Id. at 1377. It did not hold or say, however, that a brief absence during testimony that is relevant to the defendant’s theory of the case is Cronic error. In fact, the Court cautioned that it was expressing “no view on the merits of the underlying Sixth Amendment principle,” because “[a]ll that matters here, and all that should have mattered to the Sixth Circuit, is that we have not held that Cronic applies to the circumstances presented in this case.” Id. at 1378 (quotation marks omitted). Despite that caution, the dissent insists that more mattered in the Woods decision than the Supreme Court realized. What the Court failed to realize, according to the dissent, is that despite its protestations to the contrary, it was offering “valuable insight into the type of distinctions the Court may make if and when it takes such a case on direct review.” Dissenting Op. at 1241 n.ll. So what should we believe — the Supreme Court’s emphatic statement that it was expressing “no view on the merits of the underlying Sixth Amendment principle,” or the dissent’s insistence that yes it was? We choose to believe the Supreme Court meant what it said. See Mathis v. United States, 579 U.S. -, 136 S.Ct. 2243, 2254, 195 L.Ed.2d 604 (2016) (“[A] good rule of thumb for reading our decisions is that what they say and what they mean are one and the same.... ”). The dissent also relies on the decision in Williams v. Pennsylvania, 579 U.S. -, 136 S.Ct. 1899, 195 L.Ed.2d 132 (2016). Dissenting Op. at 1245-46. But that decision dealt solely with structural error involving a biased judge. It had nothing to do with a brief absence of defense counsel from the courtroom. To the extent that the dissent cites it for the proposition that structural error requires reversal, the answer is that of course it does but there was no structural error in this case. 3. The Vines Decision Except in the now-vacated panel decision in this case, we have not yet decided whether the brief absence of counsel during the presentation of testimony that directly inculpates the defendant is Cronic error. A couple of decades ago a panel of this Court did decide that the absence of defense counsel while government witnesses gave testimony that did not directly inculpate the defendant was not Cronic error. Vines v. United States, 28 F.3d 1123, 1128 (11th Cir. 1994). The case involved a two-defendant, two-day drug trial, and at 4:15 p.m. on the first day counsel for Vines left “for the remainder of the day” for some undisclosed reason. Id. at 1125. The opinion does not disclose how much of the trial day remained when counsel left, but it does reveal that during counsel’s absence, an FBI agent and another government witness testified. Id. at 1126. The FBI agent testified, among other things, about how the manner of shipping that the defendants used in that case “fit the modus operandi of contraband smugglers.” Id Vines was convicted on the conspiracy charge and acquitted on the distribution charge, and he argued on appeal from the denial of his 28 U.S.C. § 2255 motion that the absence of his counsel during the testimony of those two government witnesses was a Sixth Amendment violation that entitled him to have his conviction set aside. Id. at 1126-27. After noting that the Strickland decision applies only where counsel is present, the Court assumed, without deciding, that the absence of counsel during the taking of testimony is constitutional error. Id. at 1127-28. It addressed Vines’ argument that the absence of his counsel from the trial was not only a Sixth Amendment violation but also Cronic error giving rise to an irrebutable presumption of prejudice. Id. at 1127-28. The Court reasoned that “Cronic’s presumption of prejudice applies to only a very narrow spectrum of cases,” and concluded that Vines was not one of those rare cases. Id. at 1128 & n.8 (quotation marks omitted). In reaching that conclusion, the Court rejected Vines’ argument “that under Cronic the taking of evidence is a critical stage of trial per se,” and stated that “we decline to give birth to a rule that the taking of evidence is necessarily a critical stage of trial.” Id. at 1128. After reviewing the record, it found that “no evidence directly inculpating Vines was presented during his counsel’s absence.” Id. The holding of Vines fitted to the facts before the Court was that: “Where, as in this case, no evidence directly inculpating a defendant is presented while that defendant’s counsel is absent, we decline to hold that counsel was absent during a critical stage of trial within the meaning of Cronic.” Id. While panel decisions do not bind us when we sit en bane, we find persuasive the Vines holding that the taking of testimony or other evidence that only indirectly inculpates the defendant is not a critical stage of the trial. As the Court said there: “While trial counsel may exercise poor judgment in absenting himself or herself from a portion of a trial, such flawed judgment does not necessarily infect the entire trial.” Id. at 1129. Counsel’s absence was neither Cronic error nor some other type of structural error but instead was trial error “capable of quantitative assessment” and subject to the harmless error rule. Id. That is all that the Vines decision did hold or could hold. It did not hold — and because the facts of that case did not present the issue it could not have held— that the taking of any testimony th