Lets be sure of what nyquist said
IF you have a bandlimited waveform, you can *completely* describe that signal by taking pictures of it at twice the highest frequency of the signal. Even further, any BANDLIMITED signal can be perfectly described with only two points, with those points being twice the frequency of that waveform
Hopefully Im paraphrasing correctly
It says nothing of 44k vs 96k, except that you could accurately describe a BANDLIMITED signal up to 20khz with a sampling frequncy of 40Khz
We gotta agree where the frequency is that doesnt matter anymore
Unfortunately, gear that is pretty flat from DC to light, doesnt usually have a mode or variation that is bandlimited to something much lower, so when we say those pieces of gear that go WAY up sound better than gear that doesnt, its not a fair comparison as so much else could be involved.
We also, and heres the critical deal, must be able to bandlimit the signal
You're going to have a WAY easier and nicer time bandlimiting a signal under 22-24khz sampling at 96khz than at 44.1khz. Dan Lavry seems to say something around 60khz would be great
44.1khz, thats cutting it WAY close, but in theory, IF WE AGREE that 20khz is the upper limit, it should be doable