Description
The user guide was mostly written years ago and the code snippets aren't regularly run, I don't think.

This leads to unfortunate outcomes such as the documented behavior being explicitly wrong (#1433, possibly #1450)

It would be nice to have a runner go through code snippets in these Markdown files periodically, akin to how Jupyter notebooks are run in regular CI. I think the toolkit does something like this

Docstrings, too?