Make test output more comprehensive Running all tests in one transaction ensures the final output tells about any failing test, not just in the last transaction (i.e. test script). The price is that no traces of the tests are left in the database because we have to rollback in order to have no left-over test roles in the cluster.