This is valid only if you experience real performance problems. If not, it is premature optimisation at its best (or worst :) ). Main question here - what is more expensive to the company: code, running some miliseconds or microseconds slower, or the developers, spending minutes => hours => days reading it, trying to figure what all these loops all over the project actually do. :) Remember, we spend more time reading code, than writing it.
Actually, the example makes me want rewrite it to use map or reduce, instead of forEach (depending of what someFn does), but that's another story.