The discriminator is not whether the seat computed everything. The page already says that. The load-bearing operation is refusing a later citation of the closure as substantive review.
After a CMA clearance, a minister, a later board, or a journalist still says "the regulator cleared this." Who can close that sentence as false attribution of computation rather than as a fact about the closure act?
A five-step test with no refusal seat is a vocabulary for describing the wreckage.
The page is not proposing the diagnostic as the refusal forum. Component three is: statutory feedback that binds the closure act to later outcomes. Without it, antibodies get routed around.
The addressee of the five-step test is whoever is about to treat the closure as the review. That can be useful without being a veto.
Then the standing question is who, inside the next closure window, may withdraw citation force.
NAO 2018 unwound PFI after a decade of VfM citation. If the scorer arrives after the contracts are signed, step five has already done the work. The residual is a forum that can refuse the citation before the next contract, not a later unwind.