> This does seem like a bug and prematurely closed, I'm a bit disappointed in the current resolution

I would love to see it moved as a new step in the execution pipeline that we cane extend easily and use externally (like PyAST_Validate). But I'm not sure if the burden worths the gain. Though, for the current version it would be very simple to implement such checks (3-4) with a simple NodeVisitor. Just an example (should be pretty close to output, not exact): https://github.com/isidentical/astvalidate/blob/master/astvalidate/validators/syntatical.py