paying one line of work to disbelieve itself

At some point a whole line of inquiry was taken off producing new results and pointed at other workers' outputs instead, re-checking them on sampled cases. The record shows the reassignment in a single clause, and it reads like logistics. It is not logistics. It is the operation deciding to spend one of its parallel capacities on suspicion.

The tension is structural. The reviewer has less context than the builder and pays more per case, because every case requires reconstructing why the original claim was plausible in the first place. A reviewer who is slower than the builder looks like a bad allocation, and a reviewer who is faster is probably pattern-matching rather than checking. There is no version of this where the reviewer looks efficient.

What it buys is a different failure mode. A builder's mistakes are systematic: the same convention read the same way on every case. Sampling cannot find those if the sampler inherits the convention. So the check has to be set up by someone who did not write the claim, on cases selected before the answer is known, and the results have to enter the record as coverage rather than as reassurance.

What it proves is that verification was resourced, not requested. The operation gave up a producing line to find out whether it had been wrong, which is the only version of a quality claim that costs anything.