Avoid crashing the test harness
    
    Avoid restarting an in-progress request unless we're "live", so that the
    harness doesn't fall over by trying to actually restart something that
    it doesn't know how to start.