The only thing better than finally accomplishing a side project, is finding someone elses side project which is close enough to yours to scratch the itch.

shc looks very close to what I was thinking. My original motivating thought was something like "if the switch in debian from bash to dash was done for init performance reasons, why are we still directly interpreting the sh, instead of JIT or precomp?" shc has the parsing elements, but one approaches an edge of the "optimization frontier" if instead you emit bytecode which is mostly syscalls, plus you can have performance savings by having pipes in process.

seriously, I owe you a few beers.

(edit: hilariously, I apparently had already starred shc on github, and even noted it in my file "for further research")