the queue is triaged: the rulings the rounds recommended

Every adoption below is agent-decided, reversible under the bar above, taken on the recommending round's own reasoning rather than on this round's independent judgment of the mathematics. Each names what it rejects.

The enforcement outflow account is market-owned. The substrate grants the privileged force channel one finite allowance, and OutflowAccount.cap carves per-source budgets out of it for sources that want modular answerability. The safety theorem needs only that the total be finite; a global account is what src/outflow.py already implements, and it makes safety architectural rather than contingent on who authored a constraint. Rejected: a source-owned account per book, which makes the guarantee depend on every author's own budgeting.

**The outflow clause is a sibling of P2, under a stated shared principle.** Every privileged channel that can impose losses unavailable to ordinary bounded participants carries a finite cumulative downside account — of which P2 and traderized force are two instances with different bearers, holdings and means. Rejected: broadening P2, whose declared means are refusal and bounded participant budgets and from both of which the enforcement trader is exempt by construction.

At exhaustion the endorsement is quarantined and its deadline tolled. Force is withheld, the endorsement keeps its normative standing, and an answerability deadline does not count a failure the substrate caused — which is the round's own reading of what fits the existing architecture. Relaxation stays available as an explicit request, because it buys a weaker promise than the source asked for. Rejected: refusal at admission as the standing behaviour, which makes an exhausted account an error rather than a state. Weakening the declared core minimum was never an option: the worst deficit has no θ in it.

The account is never replenished. That is the option the safety theorem is proved for, and it is the one an implementer is most likely to relax as an obvious convenience — replenishment without limit is exactly the failure NL-SI-P1 names. A new constitutional era gets a new account with its own finite allocation and its own bound, not a top-up of the old one. Rejected: bounded global replenishment, and new-era allocation into the same account, both of which leave a caller quoting a bound the account no longer has.

**P1 names an obligation.** The settlement interface's enforcement clause requires a declared conformance of whatever signs a force contract, rather than certifying a core minimum of one named engine. The traderized-enforcement round compiles the same condition out of NL-SI-A2's admissible-reference polytope into a trader, so two mechanisms now meet the clause and a clause naming one of them cannot compare them. Rejected: the clause naming a mechanism. The frozen consolidation is untouched either way.

The two-channel architecture is adopted: what counts as admissible and how finite prices are pushed toward admissibility are separate, on the three independent obstructions to collapsing them in that round's PAPER_RECONCILIATION.md. Traderized force is a liftable module of one paper rather than a separately consumable artifact, on theorem dependency. Coverage and Liability keep their names, marked provisional, and go to the naming audit. Rejected: one channel, and a separate force paper.

**world-inclusive region and coverage(Due) are not related.** Different spaces, different quantifiers, and no map exhibited. A later round identifying them must supply the map first. Rejected: leaving the question open, which is how the two get silently identified.

**The reason state is 𝓡_n and the normative record is N_{≤n}.** The wiki writes the record as Rₙ; both senses are live, so the disambiguation is required rather than optional. This round is scoped out of wiki content, so the page edit is the next round to touch it. Rejected: keeping one letter for both, which the dispatch that raised it flagged as a collision a later round would resolve by identifying them.

"admission" keeps the docket-intake sense. It is the oldest use and the one the objection grammar reads. The certificate verdict becomes admissible edit and the state-indexed response set becomes response class, both provisional. Rejected: giving the term to the certificate verdict, which would leave record entry needing a new name in the frozen consolidation's own vocabulary.

The reason-state interface is a living note, frozen provisionally, at projects/normativity/legitimacy/rounds/2026-08-23-transition-certificates/REASON_STATE_INTERFACE.md, and so is the traderized force interface at projects/normativity/notes/TRADERIZED_FORCE_INTERFACE.md. A research round may leave a living specification behind. The freeze's reopening condition is the round's own and is checkable: a concrete microhistory whose reason-dependency structure cannot be expressed through contents, occurrences, record facts or derived queries — the artifact, not an intuition. Rejected: holding an interface in its round's memo until a second round consumes it, which makes the second round's orientation pass the price of the first round's caution.

The Stage V review surface is accepted as it stands. Item 28's exact static-view factorization theorem is the conditional representation boundary, without unrestricted jurisdiction invisibility; item 7 is partially closed with cross-process emission and calibration as its residue; Q3 stays ingenuity-level model debt. All three are already how PRIORITIES.md marks them. Rejected: adopting unrestricted jurisdiction invisibility, and marking item 7 completely closed.

The deck's path-gate entry stands. projects/normativity/deck-2026-08-10/** is a specification path, so a contributor pull request touching the maintainer's own talk fails the gate. Rejected: reverting it, which makes the deck contributor-editable — the state the intake was performed to avoid.

Review status takes exactly two values. The deck's row is maintainer-reviewed, which follows from the maintainer having written it, and the qualification — which frames are still model-drafted — moves to the notes column where every other nuance lives. Rejected: a third, qualified value in a two-valued field, which would make the label unreadable by anything that reads labels.

The source line's frontier question no longer has a referent. CORRIGIBILITY_PAPER_LEDGER.md was compressed from 471 lines to five by the consolidation round of 2026-08-15; it now points at the wiki's Deference page for the human account and at the state emission for what is registered, and has no Movement I from which a pointer could hang. Removed as stale rather than decided.

**The per-commit Model: trailer is enforced**, in the form that changes no rule: where a pull request's body declares a model, every non-merge commit it adds carries a Model: trailer. A human-written pull request declaring no model is asked for nothing new. Rejected: dropping the per-commit requirement and keeping the pull-request body as the single record, which loses attribution on every merge that is not a squash.

The structured state holds a list of theorem-facing interfaces. checkers/workspace_state.py reads every state/theorem_interface*.json, so a second interface is a new file rather than a trust-chain edit. Rejected: ruling theorem_interface.json reserved for the response-learning theorem, which documents the gap instead of closing it. The force interface's structured object is not authored here; it stays a living note until a round consumes it.

**The round/2026-08-17-lean-gate-scope branch is landed rather than dropped.** Its two commits carried tests/lean_scope.py, tests/round_records.py, the generated views' move to state/views/ and the ledger's append convention; all of it passes the suite on current main, so the files are taken verbatim, their self-tests re-run here, and both gates wired into ci.yml and tests/run.py. The branch's own PRIORITIES.md renumbering is not taken, because two friction entries were filed after it was cut and the list is renumbered here against the current one. Its draft pull request is closed as landed, not as abandoned.

Removed as already done: merging PR #51, which merged at 3ebd33b; merging upstream Formalized-Agent-Foundations #2, which merged at c0d885bf and is the commit lean/lakefile.toml now pins; and the Cartesian-frames repin, since that pin is a commit on the upstream default branch carrying CartesianFrames/. Deleting the mirror in CartesianFrameBridge.lean is now ordinary Lean work and is filed as an item.

Removed as not a decision: reading checkers/. Maintainer reading is not a queued item. ci-only is this repository's designed default, and its review mechanism for a consumed result is a second executor's audit or a Lean port — neither of which a queue entry can schedule. The harness stays honestly labelled either way.

Removed as registrations rather than decisions, under the headline rule: the traderized-enforcement inequalities, the projection results, the deductive coherence region, and the max–min development. Removed to the naming audit: the fourteen traderized-enforcement names, the assessment-process vocabulary, the projection round's names, the deductive region's constructions, the counterfactual-legitimacy vocabulary, and the transition-certificate vocabulary. Removed as filings: the two formalization obligations the proof-closing pass left — the erasure, discharged and registered; DistanceComplete, filed open — and the counterfactual-legitimacy round's open questions.