cfbb11
Make sure to test failures of all test runs ... instead of using a "last one wins" approach, which will hide errors in all tests except the last one. See also my comment in !213 for the initial report about this.