The difficulty with that is trust, if there’s any chance it will miss things by accident then I’d rather eat the slow runs.

Test runners that run new/changed tests first though are cool. Faster feedback on error but still comprehensive.