[William Sommerwerck]

The idea that upsampling increases resolution violates Shannon's
theorem and is the audio equivalent of perpetual motion.

Yes.


I suppose that some claim that up-sampling increases
resolution but I don't think that's its primary modus
operandi. It came about before we had good oversampling
digital filters, and the idea was that by up-sampling, you
could use a gentler reconstruction [sic] filter with less group
delay than the conventional brick wall at 20 kHz.

You can have a sharp filter and constant group delay. You just need
more-complex filtration.

Some upsampling uses interpolation. I've often wondered whether that affects
what we hear.