This does not accomplish what I was saying in that comment... If we were to decide to "apply statistics after converting units" (which I stated that I could make an argument for, but not that I actually thought it was the best option) that would have to be a change to how Axes.boxplot itself works, not a test that sidesteps that.

The statement about statistics plots (broadly) applying to physical units is to say I think it makes sense to have a boxplot that shows, e.g. "length in meters". This is not something we have built into our public interface (though we do have some forms of it in examples/other tests, and there is support for e.g. pint to interplay with mpl). This is even if it doesn't make sense for datetimes. It was more of a question than a final decision...

Further, there is already a PR that addresses this portion of #26864, #27038. In general our preference is to work with the first submitter of a similar PR, and to maintain authorship when a PR is built upon. I see that you commented there, though in general our preference is to at least provide the opportunity for the original submitter to respond.

Ah I see, ok. Thank you for explaining it further.

I will keep that in mind.