Diff from mypy_primer, showing the effect of this PR on open source code:

spark (https://github.com/apache/spark)
+ python/pyspark/pandas/base.py:1663: error: Unused "type: ignore" comment  [unused-ignore]

pytest (https://github.com/pytest-dev/pytest)
+ src/_pytest/monkeypatch.py:402: error: Unused "type: ignore" comment  [unused-ignore]
+ src/_pytest/monkeypatch.py:407: error: Unused "type: ignore" comment  [unused-ignore]

I don't think this really does what we'd want because mypy doesn't handle __new__ methods that return something other than a subclass of the type, and the new branch of the overload returns reversed[T] as far as mypy is concerned.

It's a tough spot to be in. We can expand what reversed will accept in it's constructor, but without mypy support it's only going to result in reversed[T]. If the actual return type is too divergent from that, that's an issue.

I'll check out the mypy-primer hits though to see what's going on and if they're actually being cleared in a reasonable way.

Related: python/mypy#15182

That said this does work as expected in pyright.

We could also use some test cases for this.

I don't think we can do this without mypy support, whether that's python/mypy#15182 or special-casing for reversed.

For example, it's true that we don't need the first overload, but without that, and with the current state of mypy, we would get reversed(list[int]) -> reversed[list[int]] from mypy. Right now, the unpacking in __new__(cls, sequence: Reversible[_T], /) -> Iterator[_T] means we get reversed[int] from mypy.

We could, with more safety, expand this to accept __reversed__ -> Iterable[T], but it's not ideal because mypy will still incorrectly infer reversed[T], which is a full iterator.

So I don't really think it's worth worrying about test cases; this is busted and we shouldn't do it right now.