But the question remains. Given Nyquist theorem and an upper limit to human ear response, what is the advantage of going over ~40kHZ? I understand you can encode more information, but decoding to higher frequencies than we can hear doesn't seem (to the novice) to be beneficial. Do we somehow 'detect' harmonics above what we can hear? Or is the extra information used to process something else differently, like the relative volumes of two slightly different instruments making a note at the same pitch? It's an interesting question IMO.

EDIT: upper ear response is presumably limited to ~20 kHz