From my experience the tests seem to break extremely easily and there is more time spent fixing them than finding actual bugs in the UI.