It was an experimental branch. Some folks reported better experiences on some hardware, but blind testing on Droid and NexusOne did not show a user visible performance difference and we opted to stick with the stock scheduler for the our 2.6.32/Froyo efforts.