Even some preliminary tests would suffice, as to see whether something is severely wrong in our setup. We are also seeing extremely inflated latencies the more workers we dispatch per second.