The AdaptationMatchingRoutine currently short-circuits its search for candidates and returns the first viable adaptation.

I don't think this is a problem, but it feels odd so I wanted to document it.

You can test this with the LiftFunctionsToImgTest. It uses this adaptor If you copy, paste, and rename that adaptor, the test will still pass.

I'm wondering if anyone else can come up with situations where this will cause problems? I am slightly uncomfortable with the idea that we could have duplicate adaptors and not know it, the way we would with non-adaptor ops. But can't think of any functional problems.