It may sometimes still only do a single run if it can't find enough difference to do another run. I do note that R4 can sometimes do less runs than R3.
My guess is that the optimise algorithm is better, meaning it can get it more right after the first pass and sometimes won't need more.
|