I had some time to look at this problem. Here's what I found:

1. Our docs do not say anything about __getstate__ or __setstate__ methods and their generation: https://docs.python.org/3/library/dataclasses.html
2. I think that this new implementation is in-line with other methods / attributes generation
3. It should not break any existing code, except cases where some invalid __getstate__ / __setstate__ method was overridden with the correct one (but, there's nothing we can do if user already has invalid code)
4. This seems like a bug-fix, so probably it should be back-ported to 3.11 as well

• Issue: Frozen dataclasses with slots cannot override __getstate__ or __setstate__ #104035