Currently the test job seems to hit the maximum CPU resource available. In order to prevent that we try to run tests on several workers.