Not quite sure what the failing test for 3.11 / pip mean. Could anybody hint at whether they are related to this PR or whether that pipeline is broken for other reasons? Thanks.

I couldn't replicate locally, the rerun timing check has always been a bit brittle but no idea why plot_scraper_broken did not seem to execute in test_error_messages. I've rerun the test 🤷

I think save_figures and docstring of matplotlib_scraper needs to be updated (even though block is never actually used). The scraper docs should be updated too: https://sphinx-gallery.github.io/dev/advanced.html#write-a-custom-image-scraper

Seems spurious, passing now

I think we have to update scraper.py and scraper docs, but otherwise LGTM!

Code-wise this is good to go. It's only documentation that could/should be improved. There are a number of places still describing block as tuple. Technically, that is still correct, because a named tuple is a tuple. But describing it as namedtuple would be more helpful.

I'm a bit tight with time now, so it would take a bit until I get around updating the docs as well. If you want to take this as is or update the docs yourself, that would also be very welcome.

Done, @larsoner feel free to merge.

Thanks @timhoffm @lucyleeow !