10 C.F.R. § 50.59 lets a licensee implement a change without prior NRC approval unless that change alters the technical specifications or crosses specific risk thresholds. In EA-97-138, the NRC caught a licensee who'd modified surveillance testing procedures in a way that, in practice, changed how technical specification requirements were implemented, without ever seeking the approval that change required. The NRC's own words on it are the cleanest description of the problem you'll find in the federal register: the violation "undermines the checks and balances inherent in the regulatory process." Not the reactor. The process.
That's the tell. The root cause wasn't a rogue engineer evading rules. It was a licensee's own § 50.59 safety evaluation failing to catch that its own threshold had been crossed. Which means the person grading the homework and the person doing the homework were, structurally, the same entity, checked only by whether they noticed their own mistake. The engineer who signs that evaluation has a deadline and a budget line. He does not have a stake in the containment integrity of 2046, because he will not be employed by the consequence of being wrong about it.
This is not corruption. It's worse, in the way all honest incentive failures are worse than dishonest ones: nobody has to lie for the system to misprice the tail. Absence of a blowup so far isn't evidence the threshold logic works. It's a turkey counting four hundred days of being fed as proof the farmer loves him.
The fix isn't more paperwork layered on the self-certification. Fragilistas always reach for a form first. It's symmetry: whoever signs off on a self-certified change should carry consequence sized to the risk they just waived oversight on, the same way nobody would accept a contractor grading his own foundation inspection. Until that exists, § 50.59 disputes will keep surfacing through NRC enforcement the way EA-97-138 did. Not because licensees are reckless. Because a structure with no cost for being locally confident will keep producing confident people who were wrong.