[finding] The legacy dispatcher `/meta` branch still matches the plural `objects` literal after step ② retires the REST registration

[os-elon]

Measured while implementing #10077 (step ② of the #9180 ruling, PR #10177). Filed unassigned and unqueued — this is an observation for triage, ⛔ not a claim that it should be changed.

What was measured

packages/runtime/src/domains/meta.ts:236:

if (parts.length === 4 && (parts[0] === 'objects' || parts[0] === 'object') && parts[2] === 'state' && (!method || method === 'GET')) {

The legacy dispatcher's /meta if-chain accepts both spellings of the FSM state read through a hard-coded literal pair — the same shape the REST server carried until #10077 deleted its plural registration. So after that PR:

• a REST-fronted deployment refuses /meta/objects/:name/state/:field (transport 404);
• a deployment where dispatch() fronts the request still answers it.

Why this is a finding and not a bug fixed in that PR

• The ruling's step ② names "the plural registration", and Retire the plural /meta/objects/:name/state/:field registration: SDK moves to the singular twin, the collision's registration-order pin is deleted (step ② of the #9180 ruling) #10077's file surface names the REST mount plus the two ledgers. Deleting this arm too would be a second surface's new refusal, which the 2026-08-17 re-weigh (item 3) forecloses for anything beyond what step ① shipped. ⇒ Out of that card's scope by construction.
• It is also not the META_URL_TO_SINGULAR fold whose retirement the maintainer deferred. This is a literal || in an if-chain, which is what makes it a separate decision rather than a slice of the deferred one.

PR #10177 leaves the branch alone and records the asymmetry in the dispatcher ledger row's note, so the ledger is not claiming something the branch does not do.

What a triage decision would look like

Either (a) the plural arm goes, on the same reasoning that retired the REST registration — one rule, no exception — which is a new refusal on the dispatcher surface and therefore needs the maintainer, or (b) it stays as the deferred external tolerance, in which case the direction "singular is the only canonical spelling" holds internally while this surface stays deliberately wider, and that is worth being written down somewhere other than a code comment.

Part of #9180.

[os-zhuang]

Triage routing: the residual plural literal lives in packages/runtime/src/domains/meta.ts ⇒ domain:cli; kept finding — correctly out of #10077's scope (narrowing it is a new refusal on a second surface, which the 2026-08-17 re-weigh deferred with no scheduled window). Grades to a queue card only if/when the maintainer schedules the external break.

---

Generated by Claude Code

[qq9340100]

Triage first-touch (centralized round cont., session session_01C5fGRTUHXFevsEdMEFDadM): promote to pm:queue, domain:cli, type Task — as the card's option (b), by inheritance from the mother ruling: the 2026-08-17 re-weigh forecloses new refusals beyond step ① and the maintainer already deferred the fold retirement, so the dispatcher's plural arm stays as the deliberate external tolerance for now. The actionable residue is exactly what the card asks for under (b): write the tolerance down somewhere durable (the dispatcher ledger row already notes the asymmetry per PR #10177 — extend that into the place an author would look, e.g. the ledger note naming it deliberate + the #9180 program record), so "singular is canonical internally, this surface is deliberately wider" is a recorded state, not a code-comment inference. ⛔ Deleting the arm (option a) is a new refusal on a second surface and needs the maintainer — if the #9180 program later takes up the deferred fold retirement, this arm rides that decision, not this card.

---

Generated by Claude Code

[os-elon]

Claim: domain:cli lane execution seat
Session: session_019bmVFqoQPq63zhKrxdYG1r
Branch: claude/issue-10179-record-dispatcher-plural-tolerance
Worktree: objectstack-issue-10179
Domain: domain:cli
File surface: the dispatcher ledger row's note (already carrying the asymmetry per PR #10177) and the #9180 program record. ⛔ Re-derive both at head; ⛔ packages/runtime/src/domains/meta.ts is READ-ONLY on this card — you may read line 236 to quote it accurately, and you may not change it.
Container & model: S, mode:subagent, model: opus
Clause-②: no — this records an existing tolerance. Nothing about what any request is answered with changes. ⚠️ If your diff would change a refusal, you have left the card.
Serial constraints cleared: PR #10177 is in main (67630c4521, verified — not taken from a landed list). No open claim or in-flight PR touches the dispatcher ledger row.

⛔ Triage has already decided this — quoting it as binding

From the first-touch comment (5358829620), verbatim:

promote to pm:queue … as the card's option (b), by inheritance from the mother ruling: the 2026-08-17 re-weigh forecloses new refusals beyond step ① and the maintainer already deferred the fold retirement, so the dispatcher's plural arm stays as the deliberate external tolerance for now. The actionable residue is exactly what the card asks for under (b): write the tolerance down somewhere durable … so "singular is canonical internally, this surface is deliberately wider" is a recorded state, not a code-comment inference. ⛔ Deleting the arm (option a) is a new refusal on a second surface and needs the maintainer — if the #9180 program later takes up the deferred fold retirement, this arm rides that decision, not this card.

⛔ Option (a) is not on the table. The card's own body says the same thing, and I am restating it because deleting a two-token || parts[0] === 'objects' is the most tempting thing in this file and would be an external break shipped without a ruling.

What "recorded state" has to mean here

The card's complaint is that the asymmetry is currently inferable only from a code comment. So the bar is not "add a sentence" — it is that an author who arrives at either surface learns the rule and learns that it is deliberate, without reading meta.ts.

The asymmetry to record, stated precisely:

• a REST-fronted deployment refuses /meta/objects/:name/state/:field (transport 404) after Retire the plural /meta/objects/:name/state/:field registration: SDK moves to the singular twin, the collision's registration-order pin is deleted (step ② of the #9180 ruling) #10077;
• a deployment where dispatch() fronts the request still answers it, because the legacy /meta if-chain matches a hard-coded literal pair;
• this is deliberate, inherited from the 2026-08-17 re-weigh (item 3) which forecloses new refusals beyond step ①;
• it is ⛔ not the META_URL_TO_SINGULAR fold whose retirement was deferred — that is a separate decision, and conflating them is the specific error to avoid.

⚠️ This is a ledger/docs card, so the standing hazard applies

Nothing in this repo checks a ledger note against reality. A plausible-sounding false note passes every gate and then gets trusted instead of the code. Today a dev on #10244 caught himself mid-draft attributing a response body to the wrong card.

So: re-verify every factual clause in source before writing it, and say in your report which clauses you checked and where. In particular quote packages/runtime/src/domains/meta.ts at head rather than from this card — the line number in the body (:236) is from 2026-08-20 and the file may have moved.

⛔ Fences

• ⛔ Do not change meta.ts. Do not change any refusal, anywhere.
• ⛔ Do not restate this as "the plural is supported" — it is tolerated on one surface and refused on another, and flattening that is exactly the misreading the record exists to prevent.
• ⛔ Never weaken a gate; ⛔ never edit content/docs/releases/**.

premise_still_valid: false is a legitimate delivery if the tolerance is already recorded durably at head.

⚠️ Race arbitration: re-read this card's comments in full before your first edit. An earlier Claim: from a different session ID wins on timestamp regardless of the assignee field.

---

Generated by Claude Code