I will add that the concerns of transient timing are actually a fallacy: given correct reconstruction, sampling is more than capable of producing a high-frequen

I will add that the concerns of transient timing are actually a fallacy: given correct reconstruction, sampling is more than capable of producing a high-frequency transient that crosses a given point at a given moment in time that's NOT simply defined by discrete samples. Reconstruction is key here, and no special technique is required: sampling and reconstruction alone will produce this 'analog' positioning of the transient anywhere along a stretch of time. The accuracy is limited by the combination of sample rate AND word length: any alteration of the sample's value will also shift the position of the transient in time. But since the 'timing' issue is a factor of reconstruction, you can improve the 'timing' of transients at 44.1K by moving from 16 to 24 bit. The positioning of samples will be a tiny bit more accurate, and that means the location of the reconstructed wave will be that much more time-accurate, since it's calculated using the known sample positions as signposts. Positioning of high frequency transients does not occur only at sample boundaries, so that alone isn't an argument for high sample rates. You can already place a transient anywhere between the sample boundaries, in any PCM digital audio system. The argument for higher sample rates is use of less annoying filters, and to some extent the better handling of borderline-supersonic frequencies. For me, the gentler filters is by far more important, and I can take or leave the 'bug killing' super-highs. I don't find 'em that musical as a rule.