Just to check: Have you already looked at the webpages I wrote on this a
while ago?

Yes, I had looked at them, thanks!

More specifically, bitfreezing lets you choose how many bits per sample to
'freeze'. Simply converting down to 16 bits nominally means losing 8 bits
per sample if you start from 24 bit. That may be too much or too little for
a given recording.

OK, so 24-16 (possibly dithered) truncation is a special case of
bitfreezing - bitfreezing being a general N-M (possibly dithered)
truncation?

If so, then yes, it probably gives you all the benefits of MQA, without
any patent/licensing issues.

Julf