After last week, I understand better how punq works in terms of type hints, and I believe it shouldn't work quite like it does. However, I worry that my current solution may be more disruptive than the bugs I've found.
As such, the first thing I want to do is put together a reproduction of the issue. I believe it requires multiple modules to fully manifest, so I'll have to add a bunch of auxiliary junk to the test directory.
I didn't actually do any of that today, because I was busy setting up more tooling. Speaking of which, here's my current status with git: At work, I managed to break every GUI I tried, so now I'm using the same three or so commands from the command line and trying not to think about it too hard. For personal stuff, the GitHub Desktop client has been fine, in the past, but it doesn't seem to have a linux version, so I went and installed hub. I'll only be using it for other people's projects, like punq. For my own, I prefer to use hg-git. For new projects that I don't feel like putting on GitHub, I'm using pijul.
Anyway, as I was saying, I was doing other things for a lot of the day, then focusing on this tooling, so I haven't yet added tests to punq to determine that there could be a bug. That's step one tomorrow, and then, I'll try fixing it. We'll see how it goes, but for now, I've let things go too late once again.