Does it matter that the judge sleeps during significant parts of a trial? What if there is a jury, can the judge sleep longer? What if the judge is alert for most of the time, and only misses some bits of evidence, through sleep or conscious inattention?
In Cesan v R  HCA 52 (6 November 2008) the High Court of Australia considered two appeals against convictions that had resulted at a trial by jury where the judge's sleeping had distracted the jury. French CJ summarised the facts (94):
"In this case the facts ... indicated a substantial failure of the judicial process. The trial judge was asleep on a number of occasions on the 11 days when evidence was being given. He slept at least once on most of those days and on some days on two or three occasions. Some of the sleep episodes, possibly between two and five, lasted from 10 to 15 minutes. Most lasted between two and 10 minutes. The judge's sleeping was accompanied by heavy breathing on a number of occasions and he snored when Cesan was giving his evidence. This was disruptive and caused Cesan to look around at the trial judge. The judge also slept through parts of [the other accused] Mas Rivadavia's evidence. The jury was distracted by the judge's sleep episodes and some of the jury members found his behaviour amusing and even emulated it."
It was sufficient for French CJ that there was an appearance of injustice. This made the likelihood of convictions irrelevant (97). The flaw in the proceedings had been fundamental and incurable. It was a miscarriage of justice by failure of the judicial process, constituted by the judge's substantial failure to maintain the necessary supervision and control of the trial (96). Therefore the proviso could not be applied.
That straightforward reasoning, which separates procedural fairness from the question of the likelihood of the convictions being justified on the evidence, was not adopted by the other members of the Court.
A joint judgment was delivered by Hayne, Crennan and Kiefel JJ, with whom Gummow J and Heydon J agreed. These reasons stress (104, 110) that established principles are being relied on. This, in effect, means that the obscure approach laid down in Weiss v R (blogged here 16 January 2006) is going to make this difficult.
And it does. The joint judgment reasons that because the jury were distracted it is not possible to conclude from the written record of the trial that guilt was proved beyond reasonable doubt (127). That non sequitur, repeated at 130, emerges from an attempt to apply the Weiss approach of perceiving the likelihood of guilt as it appears from the record of evidence but also taking into account the verdicts.
It is difficult to see why the verdicts should be of assistance to an appellate court in the task of reviewing the adequacy of the evidence, given that the jury were misled in some way – usually by a misdirection on the law, or by inadmissible evidence, or as here by judicial slumber.
It would have been preferable for Gummow J, instead of simply agreeing (107) with the joint judgment's approach to the proviso, to have taken the line he took in AK v Western Australia (blogged here 27 March 2008), reasoning that this miscarriage was substantial and there was no need to ask if the accuseds were guilty. The absence of reasons in the judge alone trial in AK is not all that far removed from the absence of juror attention to the evidence here.
In Cesan the majority took one approach to the proviso, namely that it cannot be applied where the court cannot conclude that the convictions were supported by the evidence (127). The other generally accepted reason for not applying it – procedural unfairness at trial regardless of the likelihood of guilt – provides a more straightforward reason, as French CJ demonstrated, for ordering a retrial of each appellant.