Args with a generic type will use the mother type, rather than a literal

Bug Report
Args with a generic type will use the mother type, rather than a literal

I noticed this issue while trying to chain together two lists with different types with itertools.chain.

To Reproduce

from typing import TypeVar

T = TypeVar('T')

class M:
    pass


class C1(M):
    def my_method(self) -> None:
        pass

class C2(M):
    def my_method(self) -> None:
        pass


def test(*args: T) -> T:
    return args[0]

result = test(C1(), C2())
result.my_method()

https://mypy-play.net/?mypy=latest&python=3.12&flags=strict&gist=d78c6875b87a3040126e1926b8f7ba0c

Expected Behavior
I expected to be able to run my_method, as all arguments have it defined. Instead the mother type is used. This issue can be mitigated by extracting the arguments, giving them a literal type, then unpack them into the function:

my_args: list[C1 | C2] = [C1(), C2()]
test(*my_args)

Actual Behavior

main.py:26: error: "M" has no attribute "my_method"  [attr-defined]

I believe this happens because untyped lists defined with items will use the mother type of all items. I don't think this is very intuitive for args.

Visual Studio Code's Pylance does not give an error in this situation, even with the strict flag.

Your Environment

• Mypy version used: 1.9.0
• Mypy command-line flags: no flags
• Mypy configuration options from mypy.ini (and other config files): no config
• Python version used: 3.9, 3.10, 3.11, 3.12

Start by running the supplied mypy-play reproduction with the reported Python and mypy versions, then trace generic argument inference for the test(*args: T) -> T example. Done means the mixed C1 and C2 call is inferred to a type exposing my_method without requiring the intermediate annotated list, with regression coverage for this behavior.