I am working in some probabilistic BSGS version it will increase the "probabilistic speed" but since probabilistic means drop some keys with unlikely endings or repeated patterns there is the possibility of a misshit.
First of all, let me express my respect to you for creating an excellent Keyhunt program.
Unlikely key values can be numbers in which there are several (for example, more than three) identical digits in a row? I assume this from the fact that, from the point of view of probability, it is unlikely that there will be at least one key of this kind in the puzzle.
On the other hand, the proportion of such "beautiful numbers" in the search ranges is probably no more than ~ 20%, that is, not very much. But maybe I'm wrong, and you're excluding certain numbers for some other reason.