Suggestion

I think you need to break this problem down into smaller problems. I see two requests here:
1. Make the type system understand what's going on.
2. Do something at runtime.

I think (2) is pretty much a lost cause given that it would require extensive changes to the language and that 3rd party tools (like Pydantic) can already achieve this.

Thus (1) is an interesting thing to focus on. Without (1) there is little value in (2). I tried to make inroads on this via [PEP-746](https://peps.python.org/pep-0746/). I'm not sure what the next steps are.

No files, tests, or entry points are identified. Start by reading PEP-746 and comparing the proposed type-system work with the rejected runtime request; before implementation, define a concrete change and its completion criteria.

Written by the indexing model from the issue text.