The types on stitty-documented code often do depend on the types actually being checked though. Otherwise you need to author to be diligent in their types, but if they're bad in their other docs, that seems unlikely.

You don't think that "the author thought that this function would return X, but I got Y" is useful?

Only for archeological, and maybe sometimes psychological investigations of code.

That line of reasoning doesn't make any sense to me.

If the author / last person who touched the code misunderstood it then there is a 100% chance there is a bug here.

Knowing that off the bat is quite helpful. And it also tells something about the code quality if it is a common occurrence.

Could be an indicator to look up the file in git log.

Which need would send me back to pypi to look at other prior art, if possible.

Some people are willing to be meticulous about references even when they're Less Good Than One Might Wish at picking names or explaining things.

type hints are documentation, and machine readable documentation to boot!