Load testing with random delays

The QA team needs to operate a test network for 72 hours under conditions where iptables has been configured on nodes to drop packets in a way that creates periods of patchy network brownouts, and has seen no surprising loss of consensus, data corruption, crashes, stalling of catchup, or dysfunctional view change. (This experiment does NOT need to demonstrate sustained performance of 10 trans per sec; it is okay if performance is degraded during brownouts. It also might be okay to see some loss of consensus that is NOT surprising, if a brownout is very severe--as long as consensus is restored after the brownout.)