> The extra dynamic range helps bypass bad mastering.
I'm not an audio engineer by any means, but I'm not sure that sentence makes sense. Mastering is what produces the final audio output - if you take a badly compressed master it's not going to matter if you export it to 16 or 24 bit.
>Nor can anyone hear the difference between a 16bit and a 24bit file, all else being equal.
While we don't know the exact dynamic range of human hearing, we do know it to be above 96dB.
Either way, using these 96dB in 16bit requires careful mastering.
With CDs, it is not possible to e.g. make a song quieter than the others in the CD without losing this 96dB range. There's no "ReplayGain" metadata in the discs; the format is aged and not fancy at all.
The mastering would have to be basically non-existent for it to use a dynamic range larger than available in 16bit, and would mean noone would ever hear the quieter parts.
If we're talking unmastered recordings for archival, yes 24bit is very useful. For final mastered copies released for listening it's completely untrue that 16bit doesn't offer enough dynamic range.
The way the “loudness wars” work is by reducing the dynamic range to be far narrower than what’s available at 16bit, which gives the impression of loudness even though the loudest sounds are no louder.
Lets say the mastering process compresses the recording extremely to a 42dB (8bit) range - outputting that to either 16 or 24 bit will make zero difference. The audio will still have a dynamic range of 42dB.
The extra dynamic range helps bypass bad mastering.