The Basic usage guide backport also appears to have broken things. I have half a mind to force-push these backports away, since this isn't main.

Oh, thats confusing. Sorry for the breakage, and of course feel free to revert what is needed.

... that is even more confusing, as the back ports passed CI - I guess its not worth having a different CI for these branches, but...

We do not run the link-checker as part of CI and the CI does not try to install the released version.

Given that we now have a handful of git-related checks on GHA, I suspect would could set up a check that asserts nothing in the lib or src directories are changed on *-doc branches.

I'm mixed on force-pushing to -doc. On one hand I'm against doing it as a bit of a social / process control thing. "do not force push to the matplotlib/matplotlib repo" is a very easy rule to understand and enforce. But we have broken 2x in the last 5 years (once "for real" and once kinda-sorta-language-lawyer-maybe-not) so it is actually "do not force push to the matplotlib/matplotlib repo except in extenuating circumstances" which is fuzzier. Doing it here would be another step down the path to normalizing force-pushing to matplotlib/matplotlib (slippery slope etc etc) so we should not.

On the other hand, there is no reason to keep commits + their reversions around, this is not main and I strongly suspect that @QuLogic is the only person in the world that currently has it as a local branch so just do the simple thing and move on.

I'm pretty 50/50 on this and am OK with either option.

Fixing the usage guide is less work than I thought (only one line), but now I see other stuff was merged, so I'll need to check that that is okay.

I think there was only #21813 and #21815...