@hugovk are you ok exposing this in rc2 in the FT build? Without these these APIs tools cannot add proper free threaded support when they need to have hard sync points

I am not changing the name for now so we can backport them (if @hugovk is ok with it) to 3.14 in the private headers. We can do another PR for 3.15 exposing them as unstable APIs outside private headers

For memray, I think you should be able to use _PyEval_StopTheWorld(interp), which is already exposed (instead of _PyEval_StopTheWorldAll()).

_PyEval_StopTheWorld() stops all the threads in an interpreter
_PyEval_StopTheWorldAll() stops all the threads in all interpreters

We don't use _PyEval_StopTheWorldAll much. We use it before fork() because that's process-wide, but most things including sys.setprofile() use _PyEval_StopTheWorld().

@hugovk are you ok exposing this in rc2 in the FT build? Without these these APIs tools cannot add proper free threaded support when they need to have hard sync points

Okay to backport provided someone else approves the PRs, and no-one from the C-API Workgroup objects.