Totally seconding the 80/20 rule here. It's probably less frustrating and faster for everyone if you can work on compartmentalizing failures that do slip throug

Totally seconding the 80/20 rule here. It's probably less frustrating and faster for everyone if you can work on compartmentalizing failures that do slip through or only happen for those two customers with very special data sets that you most probably wouldn't catch in testing either because your test suite won't execise the full 10TB data set anyway (assuming here because we don't know your product)