I was aware that it was theoretically possible to verify that the output was part of that subset, but I wasn't aware of merlyn's work. Thanks!

(Of course, P::RD is insanely slow, so this could use some work.)