Probably a dumb idea, but how about we split some data in some other branches to lower the size?