This is to better catch distribution errors. It's easier to notice that run-tests fails than to notice that one of many test suites is silently skipped.