By all possibilities, you mean ᶠ? I think it just finds the result of every possible choice, within the normal Prolog framework of choice points, so if the issue is with CLP(FD) then... yeah
...on a probably related topic, I thought I had a different way to golf it down but it times out on TIO, and when I try to run it locally it gives me infinite zeroes, so I hope it's because somehow one of the small changes I have locally messes it up and not because it's broken entirely by virtue of being run locally
I can't quite tell what it's doing either, but the reason that superscripted ᵘ isn't in terms of ᶠ is because it finds however many unique solutions, rather than the unique solutions out of however many solutions
It could do something like superscriptless ⁱ does, but I'm not sure it would work as well in this case, because I'm not sure how it works at all
(I'm going to fall back on my procedural language intuition here, which is probably flawed.) If ᶠ with superscript applies the call_nth/2 helper metapredicate with 'nth' being successive numbers up to the superscript, then ᵘ could be implemented by the same mechanism, but with a slightly different helper predicate, let's say call_nth_unique/2, which uses a version of the goal that recurses to exclude any call_nth_unique/2 resultss for lesser values of n.