Thanks for the post. It's good stuff and true. Wish the guy did not disable youtube comments.
Many folk unpappy about audio digitisation don't think of the digitised signals converted back to analog as piecewise continuous steps but as linearly interpolated connect-the-dots signals (just like an oscilloscope displays waveforms in non-interpolated "vector" mode). Hence the belief that well-beyond-Nyquist-sampling will make for a more accurate reproduction because when you make a connect-the-dots plot, the more dots the merrier.
The Nyquist sampling theorem is a mathematically robust one and most audio hobbyists don't appreciate how rock-solid irrefutable it is. But its advocates (including the guy in the video) also underplay the stringency of the underlying assumption that the input signal is bandwidth limited. An old vinyl record may be analog bandwidth limited to ~15 kHz, but the signal vs frequency fall-off is not infinitely steep. There is still signal there well beyond the Nyquist cut-off and that will either be lost (which is usually okay) or more likely, with sloppy AD conversion techniques, show up as aliasing artefacts in the digitised signal.
Digital is as good and better as analog when done right. But sometimes it's not done right.