I'm curious why they don't do some sort of cost-estimation when doing the regex compile. Oftentimes you use a regex only a few times in the life of the application, and it'd be nice if maybe it would build the regex automaton with some time-budget cap (say, N states created) and fallback to PCRE until its fully built.

(Backseat nitpicker here. And just thinking about it seems deliciously hairy.)