Okay, I got different filename functionality integrated into limit-coverage, and proved it out by testing it against actual code because it has literally no useful tests. I wasn't able to build pijul myself, oh well. The more I tinker with prototyping my task runner in Jupyter, the more its editing interface reminds me of the various stumbling blocks that I experienced with Sublime Text. It's not that I think that there's necessarily such a thing as an editor that manages to follow a simple algorithm that always does what I want, but I think I would be happier with Jupyter if it didn't ever auto-insert tabs, or auto-close delimiters when there's no text selection.
I think what contributes to this is that I'm writing dataclass-style classes, so I'm maybe using variable annotations more than most people who use Jupyter? And on the delimiter side, Jupyter just has the exact same janky behavior as Sublime where, if you delete from the middle of a run of similar end-delimiters, it will delete "too much" and you will end up with missing closing delimiters. (I'm not sure that's the precise behavior that triggers this condition, but it's something like it.) Maybe I should put in bugs later.