Czar,
CD is based on principles of Nyquist sampling theory.
I have a lot of experience applying that in digital image processing and resampling. It works quite well in digital imaging. I ran a lot of test cases over the years on various forms of digital imagery that convinces me of that.
Granted, it leaves little room for error, but it works pretty well when done right, even if not 100% perfectly in all cases.
That jives with what my ears tell me these days. It's a fine line between good digital and poor, but luckily the technology used with the format has improved immensely over the years and most of the problems resolved. CDs still being as relevant as they are 30 years later and the lack of traction on any scale for high res audio formats these days despite the technology to enable it existing testifies to that.
It is true that CD redbook spec has a fixed limit for dynamic range and more there is clearly possible with more bits, but the dynamic range possible already well exceeds what most consider safe levels for listening, and I have heard good CD format playback sound indistinguishable from good vinyl to me even with massed orchestral violins(though not as often as I'd like), so there you go.