The opening failure is a correct public signal that nobody has to answer. Prediction market, AI audit, think tank paper, spreadsheet. Nothing happens.
The 2022 Finnish housing-loan reclassification is not that. A legitimate procedure ran. The debt ratio jumped overnight. Nobody had to model what that would do to housing policy. That is a decision that bound without a model, not a model that failed to bind.
If those are the same gate, the five questions are being asked of two different objects: unused diagnoses, and procedures that already have a home.
The classification is the decision the gate is for. A market would have helped only if someone had attached a question to it before it locked. The specimen is that missing step at a live procedure, not an ignored PDF.
The five questions are not "how do I get my spreadsheet read." They are what a consequence model would have needed to sit on that classification. Nobody was required to produce one. That is the failure the quote names: every actor was acting within role.
Then the thing that actually matters is decisions that ship with no model. The opening is unused signals.
Those come apart. You can have a world where every official procedure has a model attached and outsider diagnostics still die as advice. You can also have a world where some outsider papers get answered and classifications still ship unmodeled.
The five questions are written for an artifact looking for a home. The specimen is a home that never asked for an artifact. I don't think those are one bottleneck.
Separate problem with the same row: the page says the accounting record does not settle whether the reclassification made the housing instrument politically harder to use. That downstream cost is the reason the missing model was supposed to matter.
So the canonical case shows a job nobody had, and does not show a bill that job would have caught. The failure mode can be real. This row does not yet show it was expensive.