Might have missed some conversations but not sure I am on board with this change. Wouldn't it be easier to just update those docstrings on our end?

@WillAyd
If I understand your proposal, I don't think it's easier for a couple of reasons. Firstly these docstrings inherited from the datetime library are not in the pandas library as a string, at least I can't grep them.
More importantly, editting them to adhere to GL01 and GL02 may seem trivial but editting them to adhere to the rest of the list of errors is not trivial and just strange. If we intend to eventually have the entire list of errors on validate_docstrings.py to be checked by the CI then it makes sense to be able to exempt inherited docstrings. For example, RT01 tests for a returns section in the docstring and then there are more tests about the returns section. (Introspection to check whether the docstring is inherited is another option that was considered but decided to be overkill).
This way presents its own problems, like having docstrings that are not very helpful and having to make sure not to add a docstring to the whitelist by mistake. Still, unless there is something I'm unaware of, it is a good way forward to get more docstring errors to be checked by CI.

Why don't we just exclude sphinx from generating doc pages for these attributes? Most of the docstrings for these don't really offer much value:

Seems simpler than adding complexity with this PR

Closing as stale though not sure this is the right approach anyway. If interested in continuing let's pick up conversation back in referenced issue