1½ Topics: realtime throttling and user-space adaptive spinning

From my understanding, simply always spinning for a short duration before making the waiting syscall should be cheap enough (assuming a good short duration is chosen) when using appropriate pause instructions. Adaptive spinning would also have to have a spin limit (as there is still no guarantee that the waited-on thread continues to run after the check or is otherwise giving up the lock anytime soon), so only on very specific situations this would possibly give a tiny advantage, while also introducing significant book-keeping overhead. Just my 2 cents.