Here is another illustration of a divergence of opinion between appellate judges over whether to apply the proviso: R v Van  SCC 22 (28 May 2009).
The proviso permits the court to dismiss an appeal if it considers that an error at trial was not substantial. The meaning of substantial is at the root of the difficulties.
The Supreme Court of Canada split 5 – 4. The minority, in a judgment delivered by newbie Cromwell J, did not discuss the law on the proviso. For the majority, LeBel J (with McLachlin CJ, Deschamps, Abella and Rothstein JJ) summarised the law (34 – 36) as requiring the Crown to establish that the error was of one of two kinds:
- An error that is harmless on its face or in its effect, in the sense of being trivial, peripheral, or actually favouring the defence. The idea here is that a jury need not be "perfectly" directed, as long as it is "properly" directed (citing Lamer CJ in R v Jacquard  1 SCR 314 at para 2). These errors are such that if they had not been made any reasonable judge or jury could not possibly have rendered a different verdict.
- An error that, although not minor or not without prejudice to the defence, occurred in a context where depriving the accused of a proper trial is justified (citing Sopinka J in R v S (PL)  1 SCR 909) and the cost and delay involved in a retrial is avoided (citing Binnie J in R v Jolivet  1 SCR 751 at para 46). This requires that the Crown case is so strong as to be overwhelming, and in assessing this any possible measure of doubt as to the strength of the prosecution case favours the appellant.
Well, permitting that second type of error is contrary to the widely accepted dictum of Lord Bingham in Randall v R  1 WLR 2237 at para 28 – mentioned in these notes as recently as 28 May 2009. The Supreme Court of Canada made no reference to fair trial requirements, and its approach is consistent with the view – repulsive to a civilised person - that a guilty accused need not be tried fairly.
Failure of the Supreme Court of Canada to update its jurisprudence on the proviso is important in this appeal because the type of error at trial was of the second kind, as is indicated by the Crown's concession (noted by Cromwell J at 97) that the error was serious.
The better approach to the proviso is to ask whether the error deprived the accused of a fair trial, that is, a trial where the law was correctly applied to facts determined impartially. If the trial was not fair, the appeal must be allowed. If the error did not prevent a fair trial, is there a real risk that it deprived the accused of a more favourable verdict? If there is such a risk, again the proviso cannot be applied.
However, even under this suggested approach, the Court in Van would probably have split on whether the trial had been fair. The minority would still say that the error may have affected the jury's assessment of credibility so that it is impossible to say whether they would have convicted in the absence of the error. The majority would still say that the error would not have affected the jury because they would have thought the witness was referring to his earlier evidence and not to inadmissible hearsay. In reasoning this way the majority sidestep the Crown's concession and treat the error as being of the first kind.
The possibility is that even the best approach to the proviso yet devised will not prevent appellate judges from coming to different conclusions when evaluating the significance of errors. But when they approach their task without a focus on trial fairness they are likely to consider errors insubstantial when the prosecution evidence is strong.