Adding scope to parametrized fixture changes test ordering, (e.g. confuses PyCharm)

The output of pytest test_.py (specifically the order of tests) changes when using a function-scoped parametrized fixture versus any other scoped fixture.

# ./test_.py
import pytest


@pytest.fixture(
    scope="class",
    params=["one", "two"],
)
def f(request):
    return request.param


def test_one(f):
    assert f


def test_two(f):
    assert f

Running the MWE with a function-scoped fixture produces this output:

❯ pytest -v test_.py
====================================================================== test session starts =======================================================================
platform darwin -- Python 3.9.6, pytest-7.3.2, pluggy-1.0.0
rootdir: /Users/cleancoder/dev/scratch/pythonProject
collected 4 items                                                                                                                                                

test_.py::test_one[one] PASSED                                                                                                                  [ 25%]
test_.py::test_one[two] PASSED                                                                                                                  [ 50%]
test_.py::test_two[one] PASSED                                                                                                                  [ 75%]
test_.py::test_two[two] PASSED                                                                                                                  [100%]

======================================================================== 4 passed in 0.01s =========================================================================

Changing the scope to class also changes the test ordering:

❯ pytest -v test_.py
====================================================================== test session starts =======================================================================
platform darwin -- Python 3.9.6, pytest-7.3.2, pluggy-1.0.0
rootdir: /Users/cleancoder/dev/scratch/pythonProject
collected 4 items                                                                                                                                                

test_.py::test_one[one] PASSED                                                                                                                  [ 25%]
test_.py::test_two[one] PASSED                                                                                                                  [ 50%]
test_.py::test_one[two] PASSED                                                                                                                  [ 75%]
test_.py::test_two[two] PASSED                                                                                                                  [100%]

======================================================================== 4 passed in 0.01s =========================================================================

The correct (aka expected) output allows PyCharm to group the test results nicely:

The other output confuses PyCharm (and also me when reading the results on the CLI directly):

Other details:

• Python 3.9.6
• macOS 13.4

Output of pip list:

❯ pip list
Package        Version
-------------- -------
exceptiongroup 1.1.1
iniconfig      2.0.0
packaging      23.1
pip            22.3.1
pluggy         1.0.0
pytest         7.3.2
setuptools     67.0.0
tomli          2.0.1
wheel          0.38.4

Start by running the minimal example in test_.py with the fixture scoped as function and class, then compare pytest -v test_.py collection order. Trace how parametrized fixture scope affects ordering and verify that the result consistently groups each test's parameter cases as shown in the expected output, without confusing PyCharm.