I'm planning to, instead of redoing the old in-code metadata thing I had, figure out a config file format and have the metadata in there. If test code wants to access the metadata, it can query the file the same way as the script will.
That all seems pretty reasonable. Because limit-coverage currently fulfills my needs well enough, my next plan is to start working on the fiddly refactors and internal documentation required to get Structured Data to support product types.
I don't think I have the focus to talk about this any further right now. I'm mostly just having conversations with myself in the issue comments. Okay, can't keep this up any longer. Good night.