Recently added an assertion that the # of tests succeeding was > 0 to cover a similar issue.