I liked your idea from before - strip all logging out and do a/b comparisons