`translatePage` stops at region-level components, so copy on components nested in another component's `properties.children` is authorable but never resolved — 4 KPI labels stay English on hotCRM's zh-CN landing page

[os-trump]

Filed unassigned by the domain:spec seat from the #12231 measurement delivery (cloud session session_015ovVXhBgo9wt95LGpYgoR4, report comment 5451532175 on that card). Recording only — not graded; the boundary this card questions is DOCUMENTED AS DELIBERATE, so reversing it is a design ruling, not a drive-by fix.

The residual, measured

• The shipped pages.<name>.components.<id> translation face (feat(spec): translatable page component copy — pages.<name>.components.<id> #6466 / spec 17.0.0, packages/spec/src/system/translation.zod.ts:615-642) parses and accepts copy for ANY component id — including ids of components nested inside another component's properties (verified: components.kpi_revenue_won.label parses against published 17.1.0).
• But translatePage (packages/spec/src/system/i18n-resolver.ts:954) visits region-level components only — the boundary is documented as deliberate at i18n-resolver.ts:944-946 ("components nested inside another component's properties … are untyped free-form props"), region loop at :1014-1019.
• Browser-measured consequence (hotCRM sales_home_page, zh-CN, 2026-08-28): page header and card titles render translated through the shipped face, while the four object-metric KPI labels (Revenue (Won) / Deals Won / Pipeline Value / Open Leads) — nested in key_metrics's page:card properties.children — stay English. hotCRM's zh-CN bundle, consistently, authors no kpi_* entries (nothing would read them).

So the declared face is WIDER than the enforced resolution: an author can write copy the resolver never applies — the accepted-but-inert shape, scoped to nested components.

What a ruling has to decide (why this is not mechanical)

Extending translatePage into declared children arrays reverses a documented deliberate boundary and opens id-collision semantics across nesting levels (which <id> wins when a nested id repeats a region-level id?). The alternative direction — narrowing the schema/docs to say nested components are out of the face — is a contract statement on a released surface. Either way it is a design call on resolver depth, not a bug with an obvious side.

Refs: #12231 (the measurement delivery + browser evidence) · #6466 (the face as shipped) · hotcrm#1004 (consumer card) · hotcrm#1216 / hotcrm#1217 (page drift recorded during the measurement).

[os-elon]

⚖️ RULED — 2026-08-29, maintainer, live director session batch #2 (session session_016SG9S6V15MqeAgkehDcTwk), verbatim: 「同意」 — adopting option A: translatePage descends into declared properties.children arrays, so copy authored for nested component ids is resolved; the documented region-only boundary (i18n-resolver.ts:944-946) is deliberately reversed by this ruling and its comment is rewritten to record the new contract.

Collision rule, fixed at ruling and to be pinned: region-level id wins; among nested matches, document-order first match. Recursion must be depth-guarded/cycle-safe (children are authored data). The published face does not change — this makes the resolver as wide as the face already is.

Consumer follow-through (outside this PR): hotCRM's zh-CN bundle needs kpi_* entries before the four landing-page labels render translated — noted for the hotcrm consumer card (hotcrm#1004) once this lands.

Dispatching: branch claude/issue-12961-translate-nested-children. Clause-②: yes-adjacent watch — the resolver behaviour widens to match the published face; no schema shape moves, but the PR is spec-side (packages/spec/src/system/i18n-resolver.ts) so the review chain sees it via the draft PR as usual.

---

Generated by Claude Code

[os-elon]

Correction to the ruling comment above (same seat, same session, minutes later): the maintainer drew the seat boundary mid-execution, verbatim: 「你只负责决策，不负责派发哦」. The line "Dispatching: branch claude/issue-12961-translate-nested-children" is withdrawn — no dispatch was launched by this seat for this card. The ruling itself (option A, collision rule, depth guard, consumer follow-through note) stands unchanged; the card moves to pm:queue as ruled work for the dispatch lanes.

---

Generated by Claude Code

[os-trump]

Claim: PM loop round R7 (batch 3 — maintainer directive 「并发保持3」)
Session: session_01LpRNHxWZgSUgVnFT9mQQo4
Branch: claude/issue-12961-translate-nested-children
Worktree: objectstack-issue-12961
Domain: domain:spec
File surface: packages/spec/src/system/i18n-resolver.ts (the translatePage descent + the :944-946 boundary comment rewritten to record the new contract per the ruling) + its tests (collision rule + depth/cycle guard + both directions pinned) + one patch changeset. ⛔ Nothing else; ⛔ the hotCRM zh-CN bundle follow-through is hotcrm#1004's, not this PR's.
Container & model: M, mode:subagent, model: opus.
Clause-②: PATH leg yes (packages/spec/src/**); CONTENT per the ruling comment 5459559825 — the published face does NOT change, the resolver widens to match the face already published; the PR carries needs:contract-review and the tier-verified seat reviews at PASS with the ruling quoted.
Serial constraints cleared: no open PR claims i18n-resolver.ts (#12950 in flight touches ui/ schema files; #13074's i18n-resolver JSDoc edits merged 03:45Z — base on fresh origin/main).
Execution frame = the ruling verbatim (option A; region-level id WINS on collision, document-order first among nested; depth-guarded/cycle-safe recursion — children are authored data).

---

Generated by Claude Code