type inference failure on custom iterable class unpacking

Bug Report

When unpacking a custom iterable class to build another iterable, mypy does not correctly infer the resulting type i.e. CustomIter[CustomIter[_T]] is equal to CustomIter[_T] (that should not be true)

To Reproduce

proof-of-concept

Expected Behavior

Nested iterable unpacking type checking works fine when using standard types (e.g. list[list[_T]]) however custom iterable classes should have the same behavior.

Actual Behavior

As presented on the playground proof-of-concept the issue is that mypy does not mark the failure on this incorrectly defined function:

def case_1(iter_of_iter: CustomIter[list[int]]) -> CustomIter[int]:
    # Mypy fails
    #
    # Mypy does not flag that the return type does not match.
    # It is `CustomIter[list[int]]` instead of `CustomIter[int]`
    return CustomIter.new(*iter_of_iter)

Your Environment

• details available on the playground example.

Start with the linked mypy-play proof of concept and reproduce the failure under its stated strict flags. Trace how unpacking a CustomIter[CustomIter[_T]] is inferred, then verify that the incorrect case is rejected while nested standard iterables retain their existing behavior.