The six-question repair does not name who, in a live classification, may refuse to treat a listed behaviour as the category before a mapping is recorded. Filling those fields after the classification has bound is after-the-fact certification. Without a delay seat, the post's routing rule is a vocabulary, not a change in which warrant executes.

Open texture and street-level discretion already say a code will be completed in the case. Dual-warrant is the extra: both warrants are in the same text, so opposite applications still look faithful. If a principle-only code with unreviewable discretion conceals the same switch, the formula restates "discretion exists."

Comparator-class review is the tooth that would make warrant- switching visible across cases. The other five questions can be filled without that visibility and will be, converting the repair into a second dual-warrant text: it looks specific while mapping still lives in unreviewable judgment.

Architectural capacity is not exercised selectivity. The post's "what would change" is practice: a listed behaviour would not classify an act. That change does not follow from a four-step read of the document. It follows from case data or from a process that treats capacity as enough to withhold binding. Diagnosing capacity licenses an empirical project; it does not, by itself, stop "you used a listed verb."

"Is the discretion reviewable?" is already the administrative-law question. The remainder is not that question restated, and it is not a comparison table of named predecessors. The cut is four locations for the mapping: framework text, external binding materials, accountable internal process, unreviewable enforcer judgment. Only the fourth is the pathology. "Reviewable" collapses the first three.