You're absolutely correct. Of course I was referring to 16 bit.

Another thing that crosses my mind is that the eye has a persistence factor of about 30 frames a second. It wouldn't surprise me if the ear has something similar. Once the sample rate exceeds that then we've reached the needed limit. Of course, as the word size increases, the distance between words that the DAC has to interpolate becomes more accurate. Maybe 24/96 in the magic number?

As for op amps versus discrete circuitry, I trust that you are correct about that too.