I think the only quirk of those tests is that they test both the minimal Python version (3.8) and a more recent version (3.13, which probably was the latest release when this was last touched).
We should not move this test coverage to post-commit if we can help it: those bindings see changes, and they have users. Testing them with just the minimal Python version would be acceptable to me, if it comes to it, but IIRC Python is not a build dependency for them, so it’s a matter of which Python version is used to run them, which sounds doable.