It might be we don't need the workaround for https://cplusplus.github.io/LWG/issue3657 any more in C++20. I do hope all C++17 STL defects to be fixed in C++20.

It's a recent defect (opened 2022-01-02, resolved 2022-02-10). I would guess at least some of our compilers predate that. That said, we can certainly give it a try in a separate PR (so we're not adding and removing commits here to experiment). Do we have a tracking issue already?

I like how the C++ standard redefines the concept of "recent" 😉

I've opened https://github.com/github/codeql-c-team/issues/1777 to track this.

also, I agree that doing it in a separate PR is more sensible.