Does Facebook use this system to turn on new features for a certain subset of the user base to test those features out, or is this mainly for A/B testing existing features?
PlanOut has been used both for evaluating the effects of new product rollouts / encouragements and for more scientific experiments that look at the effects of existing features. You may want to check out Section 4 of the PlanOut paper at:
http://facebook.com/publications/563103330455476/