I told myself back at the beginning of the summer when I was feeling awesome and on-track that surely no matter how well things seemed to be working then, that there would still likely be some setbacks. I feel comforted now a bit not having gotten too over-optimistic then.
The peculiar why-didn't-I-notice-it-before thing is that, in a certain sense, I never need the downarrow operator at all, and I never need @ at anything but base types — given a signature, I can push all the @s inward ("β-reducing" the downarrows along the way) until everything is just foralls and @s-at-basetypes. Which looks suspiciously like I'm just indexing all base types with worlds! The fortunate apparent thing is that the use of constraint domains is not "open-ended" in the sense the resource monoid is, so maybe I'm not totally fucked by the Special Case Of Prior Art Gods.