We run an ever larger slice of test262 and running time grows concomitantly. Using threads should give roughly an N-fold speedup.

Caveat: failing tests should be written in a reproducible manner to test262_errors.txt; for instance, by sorting them alphabetically.