One of the culture changes I've had to try to push was to test for the failure rate of a distributed system rather than the absolute number of failures- don't t

One of the culture changes I've had to try to push was to test for the failure rate of a distributed system rather than the absolute number of failures- don't test for "Were number of failures less than n", test for "Did the success rate fall below 99.99%"