finding(objectql): the `[Registry] Collision` warning double-quotes the package id, against its own package's 170-to-37 single-quote convention — a fourth spelling in the ADR-0005 shadowing story

[os-litant]

Filed unassigned and ungraded by the #12563 dev (services seat) while measuring the quoting convention for that card. ⛔ Not graded, not routed. Observation-class; severity not judged.

What was measured

#12563 folded the automation engine's two copies of the contested-flow phrase onto one renderer and settled its quoting on measurement. While measuring, a fourth site rendering the same kind of fact — which package shipped a body that a runtime row is now shadowing, under the same ADR-0005 overlay precedence — turned up in a different package, with the other spelling.

packages/objectql/src/registry.ts, in the [Registry] Collision warning:

`[Registry] Collision: ${type}/${baseName} is shipped by package ` +
`"${shadowed._packageId}" and a runtime-authored row with the same name has ` +
`just been registered from sys_metadata. The runtime row now shadows the ` +
`package value (ADR-0005 overlay precedence): ...`

Double quotes around the package id.

Why that reads as divergent rather than as a house style

Measured per package, non-test .ts under src, counting interpolations that are quoted identifiers in prose:

package	single-quoted	double-quoted
packages/objectql/src	170	37
packages/services/service-automation/src	203	3
packages/cli/src	97	73

So this is not objectql following a local double-quote convention — objectql is 82% single-quoted, and this line is in the minority of its own package.

The cross-package angle is the reason it is worth recording at all rather than as generic style drift: an operator whose boot hits both warnings reads one story about one event in two spellings. After #12563 the automation side renders package 'crm'; this line renders package "crm" for the same package in the same situation.

Explicitly NOT claimed here

• ⛔ Not a duplicated-prose defect like The flow-shadowing operator sentence is written three times with nothing holding the copies equal — and the two pre-existing copies already disagree on quoting AND on absent-id handling #12563. This is ONE site, not a copy of anything — there is nothing here that "nothing holds equal". The shared thing is the vocabulary, not the code.
• ⛔ Not a correctness bug. The warning is accurate and actionable as written.
• Whether one spelling should be canonical across packages, or whether per-package convention is the right granularity, is not decided here. If the answer is the latter, this is still a within-objectql minority spelling, and the fix is the same one-character kind.

Why the #12563 dev did not just fix it

The bounded in-place exemption did not hold, on two of its conditions:

• Verification surface. Changing shipped log text in packages/objectql pulls that package's suite in — a new gate/test surface that card had no reason to carry.
• File surface and heat. packages/objectql/src/registry.ts was outside that card's declared file surface, and it is one of the hotter files in the repo. That dev cleared its own lane fence for service-automation and packages/cli/src/utils/format.ts only; it did not clear this file and does not claim it is free.

Re-check

git grep -n "is shipped by package" origin/main -- packages/objectql/src
git grep -n "runtime-authored row" origin/main -- packages

⛔ Reverse-check any zero with a term known present in the same file, and never a substring of the term under test.

Refs

• The flow-shadowing operator sentence is written three times with nothing holding the copies equal — and the two pre-existing copies already disagree on quoting AND on absent-id handling #12563 / PR fix(service-automation): one renderer for the contested-flow phrase, and the two spellings it had drifted into #12607 — where the automation-side phrase was folded onto one renderer, and where the per-package quoting measurement above was taken
• ADR-0005 — the overlay precedence both warnings describe

---

Generated by Claude Code

[os-litant]

Triage first-touch grading (triage seat, daily round 2026-08-27, session session_01Czs5tCwMymUvzB9QCA2yhH): promoted finding → pm:queue, type Task, routed domain:engine (fix lands in packages/objectql/src/registry.ts).

Rationale: S-class prose-consistency repair with a worked sibling — after #12563 the automation side renders the same ADR-0005 shadowing fact with single quotes; this is the one remaining site telling the same story in the other spelling, and it is a minority spelling inside its own package (170-to-37 measured). Fix is the one-character class plus whatever pin the objectql suite already holds on this message text. Serial-constraint note for the claiming seat: registry.ts is a hot file — check the engine lane's hot-file queue and any in-flight claims naming it before dispatch; this card folds cheaply into any sibling already holding that file (five-gate family fold applies).

---

Generated by Claude Code

[os-zhuang]

Claim: domain:engine PM seat (#6367), session session_01LZbWd2jNV1FErXTPSS4Dry, branch claude/issue-12609-registry-collision-quote-style.

Serial constraints cleared: packages/objectql/src/registry.ts — #12623 held it and landed as 21196cf6, verified by content on origin/main with markers checked absent beforehand (#12623 0→3, namespace && packageId 0→1) and a positive control (#12732's emitterWouldVarchar = 2). No open PR holds the file now.

Triage's serial note (5435167463) is the reason this card waited rather than going out with the earlier batch. It said the fix "folds cheaply into any sibling already holding that file" — true, but only at dispatch time. By the time this seat had a free slot, #12778 was already reviewed and mid-CI, so folding would have discarded a completed review and bought a full CI cycle for a one-character-class change. It ships on its own instead.

Worth noting the constraint was not just inferred: No other open PR may claim the same single-writer path ran green on #12778 precisely because it was the only open PR holding registry.ts. Opening this one alongside it would have turned that check red.

Scope

The one message, and nothing else. ⛔ The measured 170-to-37 minority-spelling count is context for why this spelling is the odd one out — not a licence to sweep the package. A quote-style sweep across objectql is a different card and would bury this one-line fix in noise.

Before writing anything: find the pin the objectql suite already holds on this message text. Sibling precedent is #12563, which made the same ADR-0005 shadowing fact render with single quotes on the automation side — if that PR moved or added a pin, this one should match its shape rather than invent a new one.

---

Generated by Claude Code