I missed the HPy meeting for the 3rd time in a row :-( (I didn't attend it yet.)
While i know that it's different approach than HPy, I wanted to share my PEP: https://www.python.org/dev/peps/pep-0620/
It's my concrete proposal to "enhance" the CAPI but limit the breakages.
The meat of the PEP is the process to introduce incompatible changes in the C API. It's based on coordination between CPython core devs and maintainers of "popular" extension modules: https://www.python.org/dev/peps/pep-0620/#process-to-reduce-the-number-of-br...
The PEP only solves a few issues which HPy tries to fix, but my overall goal is to move the C API "towards" HPy.
As you can see, C API changes are limited by PyPI packages (backward compatibility).
My personal goal would be to disallow to dereference a PyObject*: make PyObject structure opaque. But i'm not sure that it's doable :-( at least, not "in the short term".
So far, the feedback on python-dev is not great. People don't see any benefit of these incompatible changes.
Night gathers, and now my watch begins. It shall not end until my death.