This shouldn't be necessary for perfect testsuites, but it's not acceptable that a flaw in a test causes failures in other project's test runs.