It seems to me that if nothing ever fails short of the 100% test its probably overengineered, since the 100% already includes a significant safety margin.

Alternatively if nothing ever fails the test, then they could also just stop doing these tests, since the analysis would already be shown to be sufficient.