[tests] linkcheck builder unit tests fail sporadically when run using unpredictable orderings.

[jayaddison]

Describe the bug

Although the linkcheck builder unit tests pass fairly reliably in the GitHub Actions continuous integration we have configured, they are less reliable when run using a random unit test ordering.

Ideally unit tests should be able to run independently of each other; as reported in #11285, we do have some side-effects within our test suite, and it is possible that side-effects are a factor here too.

How to Reproduce

Using the pytest-randomly plugin is a straightforward way to replicate this problem.

From a checkout of bf0bec3 I'm able to see the problem occur fairly reliably with pytest==8.1.1 and pytest-randomly==3.15.0 by running:

$ pytest --randomly-seed=16 tests/test_builders/test_build_linkcheck.py
...
=========================== short test summary info ============================
FAILED tests/test_builders/test_build_linkcheck.py::test_too_many_requests_retry_after_HTTP_date[GMT-3] - assert '127.0.0.1 - .../1.1" 200 -\n' == '127.0.0.1 - .../1.1" 200 -\n'
========================= 1 failed, 38 passed in 1.59s =========================

Environment Information

Platform:              linux; (Linux-6.6.15-rt-amd64-x86_64-with-glibc2.37)
Python version:        3.11.8 (main, Feb  7 2024, 21:52:08) [GCC 13.2.0])
Python implementation: CPython
Sphinx version:        7.3.0+/bf0bec3b4
Docutils version:      0.20.1
Jinja2 version:        3.1.3
Pygments version:      2.17.2

Sphinx extensions

N/A

Additional context

No response