Is it ok that tests fail in stable releases?