SevenCubed - 3:45 pm on Aug 21, 2012 (gmt 0)
Here's something else worth considering. Why is it that their results were so good years ago but trash now? Probably because their earlier algorithm had a mostly uninterrupted logical flow to it whereas now with every change they are probably writing exceptions to handle them. So now they probably have a mountain of objects colliding and conflicting with each other. It wouldn't surprise me if some of them even enter into endless loops with each other in ways the QA team didn't test for prelaunch. It only becomes apparent after it is set loose in the wild.