We use a simple test-runner helper that runs each test in a separate process and reports the pass/fail rate at the end.
Kristian Høgsberg