The idea is that if your master repo has no branches, part of the server side computation they have to do for pulls is not needed. Stu said they probably would not open source the server side changes that speed up git pull.