I saw similar discussions over and over again , and 'pro-hires audio pipeline' ppls do same and same mistake again - they trying to appeal to some specific issues from the deep and grim signal processing world ... while their opponents clearly demonstrates that they even do not understand basics :
Let's get back to Nyquist frequency / criteria , Nyquist–Shannon theorem etc : what u should realize at first ? ( unfortunately , this simple thing usually missing or even misinterpreting in lots of introductory papers/articles/books ) Yes, it's about sufficient condition, but it's sufficient only in case of ideal/perfect signal ( in 'math' meaning ! yes, yes, I know this is not 100% correct terms use , but it's correct enough for short note in such discussion ) . Add imperfect world where this should be implemented .... and we can easily postulate - related criteria/frequency/whatever , actually , is not about 'sufficiency' but about 'necessity' in case of 'imperfect' ( again, in 'math' meaning) signal .
Note: if u don't trust such quick and simple explanation , just try to dig more on issue, and u quickly find papers where it's explained in more 'academic' way.
And only after realising that simple fact, we can dive into , again, grim details of signal processing, conversion and 'generation' , real-word application, 'tricks' used by different software and hardware implementations etc etc .
So, no , "The sufficiency of 16-bits and 44.1 kHz has been proven by hard research data again and again" is completely wrong ... Only something like "The necessity of 16-bits and 44.1 kHz has been proven by hard research data again and again" can be considered somewhat 'close' to reality.