Our tests currently ensure that warnings are triggered whenever appropriate, but to ensure end users get a correct and meaningful warning message, we should also make sure the correct warning message is surfaced whenever we check for a warning.

And so I propose we start enforcing the usage of the match argument in all uses of tm.assert_produces_warning.

Using the regex assert_produces_warning\([a-zA-Z]*\), I was able to find around 200 bare uses of assert_produces_warning, and some of them are checking that warnings are not raised, so this will be an easy task to finish and would be a good way for new contributors to get familiar with the pandas codebase.

If all agree, then we can open a new issue which would have instructions for how to find the bare uses, and maybe we can enforce this using our linter. We could also implement a method tm.external_warning_produced similiar to tm.external_error_raised, which would be equivalent to using tm.assert_produces_warning(warning, match=None)

See also #23922, #30999 and #37261.