All right, I've got a prototype of the successor to Divide and Cover. I'm calling it limit-cover for now, and I'm thinking I'll just test it directly on Structured Data, rather than getting it set up in PyPI and such first.
I just confirmed that Structured Data's tests don't customize the module discovery any, so I don't need to worry about those features yet.
And, something is broken, but I'm too tired to figure out what. Better sleep on it. Good night.