Integration tests depend on external state, so sometimes they can be flaky if there is some situation that isn't being accounted for. The purpose of this script is to repeatedly run tests over and over in mulitple processes, waiting for failures. Logs for failed test runs are kept, logs for successful runs are discarded.