consider this PR the release candidate for the next release.

Comment here if any serious issues are found.

Based on this site's thorough tests:

https://nextchessmove.com/dev-builds

There was a statistically significant drop of 30 ELO in strength between 220207 and 220217. I assumed this would get corrected before a full release. Any comments?

yes, that was discussed quite in detail, see #3937 basically the same patch that was a significant gain at VLTC (roughly 1s per move and more), regressed in very fast games. A large number of patches/tests were afterwards tried on fishtest to improve simultaneously at STC and LTC but nothing suitable was found. This was known at the time of merging, and the decision was to merge it as using >1s per move is typical usage.

OK — great! As long as you're aware of it and it's been discuss.

merged with e6e324e

I thought it would be interesting to measure current SF15 Classical strength versus the last SF version before NNUE was introduced. The results are:
Elo: 11.81 +-1.8 (95%) LOS: 100.0%
Total: 40000 W: 11072 L: 9713 D: 19215
Ptnml(0-2): 222, 4167, 9870, 5512, 229
nElo: 22.12 +-3.4 (95%) PairsRatio: 1.31
https://tests.stockfishchess.org/tests/view/625ddddce7b02dbe28c999b9
Since Classical is sometimes still used (ex. exhibition matches at CCC) hopefully this datapoint will be useful to switch from the old SF to the latest just with NNUE=false set instead.