Inconsistent "Incompatible return type" error when `List[int]` is returned for `List[Union[str, int]]` when layered up with Tuple/Union

Bug Report

When layering up some Union/Tuple checks, at some point the type checking suddenly gets a lot more strict and starts failing, as if it's unable to fully parse the type checks anymore.

It's possibly related to this, but it doesn't explain why the behaviour changes only once it's wrapped under other things: https://github.com/python/mypy/issues/3351

To Reproduce

Here's a minimal reproducible example that shows how it works fine until it suddenly doesn't.

from typing import Union, Tuple, List

# Just to reduce the noise
S = Union[int, str]
L = List[S]

# It works with `([int],)`
def func_works1() -> Tuple[L]:
    return [1],
    
# It works with `([int, str],)`
def func_works2() -> Tuple[L]:
    return [1, 'a'],

# It works with `int | [int]`
def func_works3() -> Union[S, L]:
    return [1]

# It works with `(int,) | ([int],)` if returning (int,)
def func_works4() -> Union[Tuple[S], Tuple[L]]:
    return 1,

# It fails with `(int,) | ([int],)` if returning ([int],)
def func_fail1() -> Union[Tuple[S], Tuple[L]]:
    return [1],
main.py:21: error: Incompatible return value type (got "tuple[list[int]]", expected "tuple[int | str] | tuple[list[int | str]]")  [return-value]

# It fails with `(int,) | ([int],)` if returning ([int, str],)
def func_fail2() -> Union[Tuple[S], Tuple[L]]:
    return [1, 'a'],
main.py:25: error: Incompatible return value type (got "tuple[list[object]]", expected "tuple[int | str] | tuple[list[int | str]]")  [return-value]

https://mypy-play.net/?mypy=latest&python=3.12&gist=9477b2d4dab01919bef1aa6d2911f648

Expected Behavior

It passes the test.

Actual Behavior

In the code above. It complains that List[int] is not a valid List[Union[int | str]]. According to the issue linked above, this may not be wrong, but the behaviour is inconsistent as to when it flags it as wrong.

Your Environment

• Mypy version used: 1.11.2
• Mypy command-line flags: None
• Mypy configuration options from mypy.ini (and other config files): I just run py -m mypy without any config files.
• Python version used: 3.7, 3.11

Edit
This is nothing to do with the bug report, but in case anyone finds this issue with a similar use case, this is the fully working solution I've ended up with:

S = Union[int, str]
T = TypeVar('T', S, List[S])
FuncType = Tuple[T, T]

def func() -> FuncType:
   return a, b

Start by running the minimal reproducer with mypy 1.11.2 and compare the unwrapped Union/Tuple/List cases with func_fail1 and func_fail2. Read issue #3351 for related behavior, then trace the type-checking path for nested Union and Tuple expressions; done means the reported cases pass consistently with a regression test.