OK, I think I've made enough changes to the regression.py script to make it much more optimal when it comes to cloning (should only clone once, ever) and I've resolved the problems with cleaning sources that kept Windows testers from progressing.

Please try again ASAP.