623 So.2d 1205 (1993)
Carl N. LEWIS, Appellant,
v.
STATE of Florida, Appellee.
No. 92-0433.
District Court of Appeal of Florida, Fourth District.
August 25, 1993.
Rehearing or Certification and Rehearing Denied October 12, 1993.
Richard L. Jorandby, Public Defender, and Paul E. Petillo, Asst. Public Defender, West Palm Beach, for appellant.
Robert A. Butterworth, Atty. Gen., Tallahassee, and Michelle Smith, Asst. Atty. Gen., West Palm Beach, for appellee.
Rehearing or Certification and Rehearing En Banc Denied October 12, 1993.
PER CURIAM.
This appeal emanates from a robbery trial, and centers on a trial judge prohibiting the defense from questioning the victim about certain unconnected criminal charges filed against that victim. Also involved is an erroneous instruction to the jury on flight and the failure to conduct a Nelson hearing. We reverse on all three grounds. We have considered point two en banc for purposes of receding from Young v. State, 601 So.2d 636 (Fla. 4th DCA), rev. denied, 613 So.2d 13 (Fla. 1992).
In Florida, the law is clear that "[w]hen charges are pending against a prosecution witness at the time he testifies, the defense is entitled to bring this fact to the jury's attention to show bias, motive or self-interest." Torres-Arboledo v. State, 524 So.2d 403, 408 (Fla.), cert. denied, 488 U.S. 901, 109 S.Ct. 250, 102 L.Ed.2d 239 (1988). See also Fulton v. State, 335 So.2d 280 (Fla. 1976). Interestingly, there is no Florida case addressing the particular fact pattern now before us; that is, whether the state witness can be cross-examined about charges against him if they are not pending but have been nolle prossed prior to trial. In Delaware v. Van Arsdall, 475 U.S. 673, 106 S.Ct. 1431, 89 L.Ed.2d 674 (1986), the Supreme Court of the United States held it was error to deny cross-examination of a state witness where the charges against the witness were not pending at the time of the defendant's trial, having been dropped in exchange for the witness' testimony. However, the court held the error is subject to a harmless error analysis. Van Arsdall is distinguishable from the case at bar since there was testimony in that case of an agreement between the state witness and the authorities to drop the pending charges in exchange for the testimony. Here, there is no such evidence. However, at least one state court has rejected the notion that evidence of an agreement between the testifying party and the authorities is necessary in order to allow cross-examination. See Brown v. State, 74 Md. App. 414, 538 A.2d 317 (1988).
Applying Van Arsdall and Brown to the facts of this case, we reverse. The trial court's prohibition on cross-examination of the victim was harmful. The victim was the only eyewitness to the robbery and he was the only witness who provided direct testimony incriminating the defendant. Although a police officer provided some independent corroboration of defendant's guilt by testifying that he saw the latter drop a cardboard box as he was running down a nearby street, away from the crime scene, that officer was not an eyewitness to the robbery. The victim testified that the defendant had him place the money in a cardboard box. Further, two defense witnesses offered conflicting evidence by testifying that they saw a brown Malibu flee the scene immediately after the robbery. The defendant was apprehended in a Chrysler. Finally, there was no physical evidence tying the defendant to the crime. All of the foregoing makes the victim's testimony, and credibility, crucial. As such, the error in this case was harmful.
Defendant's second point is that the trial court reversibly erred by giving the standard flight instruction over defense counsel's objection. In Fenelon v. State, 594 So.2d 292 (Fla. 1992), the supreme court disapproved the giving of the flight instruction on the ground that it constitutes an impermissible comment upon the evidence. The court held that "henceforth the flight instruction should not be given." Id. at 295.
Although the language in Fenelon indicates that it was intended to apply prospectively only, defendant argues it applies here because this is a "pipeline" case. In Smith v. State, 598 So.2d 1063, 1066 (Fla. 1992), the supreme court clarified its position on "pipeline" cases, holding that any of its decisions in criminal cases "announcing a new rule of law, or merely applying an established rule of law to a new or different factual situation, must be given retroactive application by the courts of this state in every case pending on direct review or not yet final."
The state argues that Smith does not make Fenelon applicable and cites Young v. State, 601 So.2d 636 (Fla. 4th DCA), rev. denied, 613 So.2d 13 (Fla. 1992), in support of his argument. In Young, this court's majority, in dicta, said "Fenelon is not retroactive." Id. at 638. Judge Warner dissented from the conclusion of the majority on another issue, and also pointed out that Smith does make Fenelon retroactive.
We hereby recede from the dicta in the majority opinion of this court in Young, and hold that Fenelon should be retroactively applied on account of Smith. Other districts have held that Fenelon is retroactive on account of Smith. See Keys v. State, 606 So.2d 669 (Fla. 1st DCA 1992) (retrospective application of Fenelon is required by Smith); Dupree v. State, 615 So.2d 713 (Fla. 1st DCA 1993); Bryant v. State, 602 So.2d 966 (Fla. 3d DCA 1992) (Fenelon applies retroactively because of Smith); Viniegra v. State, 604 So.2d 863 (Fla. 3d DCA 1992).
The state further argues that even if Fenelon does apply, the giving of the jury instruction is subject to the harmless error test. The supreme court upheld Fenelon's conviction on the ground that the giving of the instruction in his case was harmless error. As previously noted, the testimony here was somewhat conflicting. While there was a positive identification of appellant by the victim and testimony by a police officer that he saw appellant drop a cardboard box as he ran, there was also testimony from other witnesses who saw a brown Malibu flee the crime scene (defendant was apprehended in a Chrysler). Moreover, there was no other physical evidence tying appellant to the crime. As such, the error was not harmless.
Defendant's third point is that the trial court failed to conduct a proper Nelson inquiry after defendant indicated his desire to discharge his counsel. See Nelson v. State, 274 So.2d 256 (Fla. 4th DCA 1973), approved, Hardwick v. State, 521 So.2d 1071 (Fla.), cert denied, 488 U.S. 871, 109 S.Ct. 185, 102 L.Ed.2d 154 (1988). The defendant concedes that the trial court in this case satisfied the first prong of Nelson by making a sufficient inquiry into defense counsel's competency; however, he argues that the judge failed to satisfy the second prong by not advising the defendant that if he discharged his original counsel, the court would not be required thereafter to appoint a substitute. A review of the record shows that the defendant is correct in his assertion; while the trial judge looked into counsel's competency he failed to advise the defendant of the consequences of discharging his original counsel as required. The failure to advise is error. See Taylor v. State, 557 So.2d 138 (Fla. 1st DCA 1990). However, this type of error is also subject to the harmless error test. In Taylor, the court concluded that it was unable to view the error as harmless because the state had not argued such. Here, too, the state fails to present any argument that it was harmless. Ciccarelli v. State, 531 So.2d 129 (Fla. 1988); State v. Di-Guilio, 491 So.2d 1129 (Fla. 1986). Thus, this error is also reversible.
We find no reversible error in the remaining points on appeal.
AFFIRMED IN PART; REVERSED IN PART AND REMANDED.
DELL, C.J., and ANSTEAD, HERSEY, GLICKSTEIN, WARNER, POLEN and KLEIN, JJ., concur.
STONE, J., dissents with opinion.
FARMER, J., dissents with opinion with which GUNTHER, J., concurs.
STONE, Judge, concurring in part and dissenting in part.
I would affirm. In my judgment, the flight instruction is harmless error. I note that only that issue is considered en banc.
I concur in our resolving the Fenelon-Smith issue en banc. If Smith left any room for potential exceptions to the policy announced, I would hesitate to reject the clear mandate of the supreme court in Fenelon. However, given the supreme court's apparent rejection in Smith of the argument posed by Justice Grimes' concurring opinion in that case, to apply the prospectivity language of Fenelon here would amount to rejecting the command of Smith. By Smith, the supreme court appears to have clearly relinquished the authority it otherwise had to determine that an opinion should only be applied prospectively.
FARMER, Judge, dissenting.
I dissent from the decision to consider this case by the entire court. We should "en banc" a case for uniformity purposes only when the prior decision in question is irreconcilably in conflict. Such decisional disharmony arises only when the earlier decision necessarily and directly confronted and decided the subject issue. Statements of legal principles in the opinion not directly necessary to the decision i.e., obiter dicta should not be enough to create conflict.
The felicity of this view was gently explained in New Port Largo Inc. v. Monroe County, 985 F.2d 1488 (11th Cir.1993) (Edmondson, J., specially concurring). As Judge Edmondson said:
For good or for bad, opinion-writing judges unlike legislators can make cases decide no more than the cases present. For example, no matter how often or how plainly a judicial panel may put in its opinions the "we hold X," "X" is not law and is not binding on later panels unless "X" was squarely presented by the facts of the case and was a proposition that absolutely must have been decided to decide the concrete case then before the court. All readers of judicial opinions, but especially members of later judicial panels, have the right (for judges, the duty) to look beyond the words of an opinion in which a panel attempts to explain the reasons for its decision and to determine what was decided in reality. This process is, I think, largely what the study of law is all about. And * * * the idea that a case is authority only for what it actually decides * * * is not "odd," at least in the English-speaking world with its common law tradition. Nor do I believe it to be odd to recognize that some judicial decisions are correct even when the stated explanation for the decision is flawed. We are bound by the decisions of earlier panels of this court: like cases should have like results. We are not necessarily bound by the words of their opinions, although we may be persuaded by them even if not bound.
985 F.2d at 1500-1501, n. 7. He then demonstrated how an appellate court like ours should view prior precedent for internal conflict purposes and potential en banc review:
Rehearings en banc are costly, timeconsuming affairs for litigants and for the court as an institution. Delay in final disposition is inevitable. The court ought not go en banc lightly. We have a high duty, I think, to reconcile decisions if possible before going en banc to resolve "conflicts," that is "to secure or maintain uniformity of [our] decisions." * * * This reconciliation is aided considerably by ignoring dicta and reducing each decided case to its exact holding as demonstrated by the actual disposition of the case given its facts. When we read cases as if every pronouncement in every opinion is law, the courts appear to produce many generalized "principles" that arise apart from the process of deciding an actual case; and there will be more errors, more conflicts, and more "law" about which various judges can disagree and think en banc hearing may be needed.
985 F.2d at 1501, n. 8.
I believe in this essential wisdom. We are now a court of 12 judges. Our court disposes of more than 300 cases yearly per judge. If the judges of this court are going to find en banc conflict in everything written in our dispositions regardless of whether the statement is necessary to the disposition, we shall soon find ourselves with little else to do but attempting to make what we publish uniform. Then we shall have little time for our primary review function. We simply cannot afford to spend the time as a full court trying to harmonize every unnecessary statement in our decisions and still do justice.
I have just reread our decision in Young v. State, 601 So.2d 636 (Fla. 4th DCA), rev. denied, 613 So.2d 13 (Fla. 1992).[1] While I concede that we did say that "the Fenelon opinion is not retroactive," 601 So.2d at 638, it is plain beyond quibble that our real holding on the subject was that there was insufficient evidence to support any jury instruction on flight. Moreover, upon concluding that it was error to give the instruction we also concluded that the error was harmless because of evidence in the record establishing guilt beyond any reasonable doubt.
In short, we held that it was error to give the instruction, which just happens to be the same result reached in Fenelon. Hence, even assuming that an error deemed harmless can create decisional conflict or lack of uniformity, here we have the strongest possible reason to disregard the erroneous statement of Fenelon's retroactivity. Fenelon would not have changed the result in Young. What difference could it possibly have made whether Fenelon was properly deemed retroactive? If I were but serving on a panel on this case rather than sitting en banc, I would have unhesitatingly signed an opinion that explained how limited the Young holding was and stated that our decision in this case does not create any disharmony.
On the merits, I do not agree with the reversal as to the prohibition on cross examination of the key state witness. On the day of jury selection, the state moved in limine to exclude any cross examination of the witness Green as to criminal charges arising from incidents connected with the subject robbery having been filed against him and the later dropping of such charges. The trial court granted the motion.
On the next day when the evidence portion of the trial began, the witness was called to testify by the state. At no time during cross examination did defendant's counsel ask the court to revisit the exclusion of the subject from cross examination or proffer questions relating to the charges or the later nolle prosequi. Indeed counsel never again raised the subject.
It is unquestionably true that a trial court's decision to deny a pretrial motion in limine is not enough to preserve an objection to the admission of evidence for appeal; it is still necessary to raise the objection again at trial. Correll v. State, 523 So.2d 562 (Fla.), cert. denied, 488 U.S. 871, 109 S.Ct. 183, 102 L.Ed.2d 152 (1988); Phillips v. State, 476 So.2d 194 (Fla. 1985), cert. denied, ___ U.S. ___, 113 S.Ct. 3005, 125 L.Ed.2d 697 (1993); Crespo v. State, 379 So.2d 191 (Fla. 4th DCA), cert. denied, 388 So.2d 1111 (Fla. 1980); German v. State, 379 So.2d 1013 (Fla. 4th DCA), cert. denied, 388 So.2d 1113 (Fla. 1980); cf. Holmes v. Mernah, 427 So.2d 378 (Fla. 4th DCA 1983) (motion in limine denied immediately before witness testified and thus no waiver from lack of subsequent objection).
When the coin is turned over i.e., when a pretrial motion in limine is granted but the proponent of the evidence fails to raise the subject again at trial or make a proffer (outside of the presence of the jury), the only extant Florida authority holds that the issue has not been waived and has been preserved for appellate review. Bender v. State, 472 So.2d 1370 (Fla. 3d DCA 1985). The expressed rationale of Bender is that the proponent of the excluded evidence has justifiably relied on the trial court's pretrial ruling and is therefore excused from actually raising the subject at trial when the witness is testifying.
I do not understand how the stated reason in Bender necessarily supports the conclusion. After all, the unsuccessful pretrial movant in limine has no less relied on the trial judge's ruling denying the motion. Both have bowed to the expressed will of the court. Neither has attempted to ruffle the judicial feathers by asking that the matter be hashed over again during trial; neither has sullied the minds of the jury by speaking the unmentionable in their presence. Why should the results be any different if reliance is the test?
The obvious answer to me is that there should be no difference because of the reasons underlying the rule requiring that the subject be raised contemporaneously during trial. As I have recently observed:
"The contemporaneous objection rule is not a mere mechanical formality which appellate judges may overlook in an effort to do substantial justice. It serves the pro-phylactic function of requiring a renewal of the objection in order that the trial judge can reconsider the earlier ruling in limine in light of the evidence since adduced and while the witness is present in court to testify as to any factual inquiry made necessary by the circumstances. Hence, it is not a question of the temporal span between the ruling in limine and the actual presentation of the evidence sought to be excluded, so much as it is a recognition of what has transpired between the two events."
Gilliam v. State, 602 So.2d 986 (Fla. 4th DCA 1992) (Farmer, J., specially concurring). To the above reasons, I add that if the trial record is silent on the subject there is no way of telling whether trial counsel had simply made the tactical decision not to make the proposed inquiry even if the court would permit it.
I thus disagree with the third district in Bender and would require all pretrial evidentiary rulings to be raised again during trial. I do not suggest that counsel blurt out a question in front of a jury when a judge has made a pretrial ruling prohibiting such an inquiry. The subject can be brought up outside the presence of the jury. When it is thus done, the court will then have a record of actual trial evidence to inform the decision; and we will have the certainty both that counsel had not abandoned the area of inquiry, and the court had remained steadfast in its determination.
While I thus dissent from the decision to reverse on the issue relating to the prohibition on cross examination, I agree with the result because the Fenelon error was not harmless and because of the failure to conduct the Nelson inquiry.
NOTES
[1] This court has a practice of circulating to every judge of this court all panel decisions on which an opinion has been written during the week immediately preceding the release of the decision. I recall that when Young was circulated, I agreed with Judge Warner on the Fenelon issue because of Smith v. State, 598 So.2d 1063 (Fla. 1992). I chose not to request a poll of the judges under our internal operating procedures for en banc consideration because I regarded the statement in the majority opinion as to the retroactivity of Fenelon as entirely inconsequential. I have learned of nothing since then to change that view.