[finding] The whole report.editor.* namespace — 106 keys × 10 packs — has no consumer left; the report editor it labelled became the spec-driven inspector #4145

Measured while implementing #4118 (PR #4137), not fixed there — that card's fence is the seven wrong-slot call sites, and this is a namespace-liveness question, not a wrong-copy defect. Filed so the measurement does not stay buried in a PR.

The measurement

report.editor in packages/i18n/src/locales/en.ts spans lines 1052–1165 and holds 106 leaf keys (breadcrumb, basic, title, titlePlaceholder, description, type, typeTabular, data, objectName, limit, columns, filters, groupBy, rows, values, grouping, dateGranularity, chart, chartType, …). All ten packs carry the set (all-locales-key-parity owns that), so it is roughly 1060 translated strings.

Repo-wide, the number of call sites that read any of them is zero:

$ grep -rn "report\.editor\." packages/ apps/ examples/ scripts/ | grep -v packages/i18n
packages/app-shell/src/views/ReportConfigPanel.tsx        (comments only, after #4137)
packages/app-shell/src/views/__tests__/ReportConfigPanel.panelTitle.test.tsx
packages/components/CHANGELOG.md
packages/plugin-report/CHANGELOG.md

Not a dynamic-key blind spot either: there is no t(\report.editor.${…}`)anywhere, and no JSON/MDX metadata references the namespace.check-i18n-call-site-keys.mjs` reports 62 dynamic-key sites repo-wide and none of them are in this family.

Spot-checked individually, each returns 0 non-test, non-locale call sites: basic, description, type, data, columns, filters, groupBy, chart, breadcrumb.

Why it went dormant

The namespace was built for the hand-rolled report editor form — the CHANGELOG entries read "All editor labels and hints are i18n-driven (report.editor.*)". That form is gone: ReportConfigPanel's body is now ReportDefaultInspector, a spec-driven inspector whose labels come from the report spec's own metadata, not from this pack namespace. ReportConfigPanel.tsx:55 still calls its remaining catalog "Legacy field catalog for the pre-9.0 object-bound report editor."

Until #4137 the namespace had exactly one live reader — and it was the bug #4118 is about: the panel borrowing report.editor.title (the Title field's label) to name itself. Moving that slot to a purpose-built report.editor.panelTitle left panelTitle as the namespace's only consumed key and the other 106 with nothing reading them.

Why no gate sees it

Every i18n gate in the repo runs call site → key, never key → call site:

• scripts/check-i18n-call-site-keys.mjs asks whether each call site's key resolves in en. A key with no call site is not a call site, so it is never visited.
• all-locales-key-parity compares the ten packs' key sets to each other. 106 keys present in all ten packs is exactly what it wants.
• scripts/check-i18n-en-drift.mjs only fires when an en value changes. These values are static.

So the packs can accumulate dead namespaces indefinitely with every gate green — the mirror image of #3810 / #4117 (dead fallbacks at live call sites); this is live keys with dead call sites.

Impact — observation class, nothing a user hits today

No user reads a wrong string because of this; the keys simply never render. The cost is carrying cost and misdirection:

• ~1060 strings across ten packs that translators maintain and reviewers diff for nothing.
• The next agent reasoning "what does this key mean" from pack neighbours reads a namespace that describes a UI that no longer exists — which is precisely the inference chain that produced The report config panel is titled "Title" and the empty state says "Object Not Found" — call sites asking for a key whose value is not the copy for that slot #4118's defect.

Suggested direction (not a ruling)

Two candidates, and the choice needs a maintainer:

1. Retire the namespace — if the spec-driven inspector is the settled direction, report.editor.* minus panelTitle is dead weight; delete from all ten packs.
2. Keep it deliberately — if the hand-rolled editor is meant to come back, or some downstream consumer outside this repo reads the packs, retire nothing and record why.

Worth deciding as a class rather than one namespace: a "no key without a reader" gate (key → call site, allowlisted for deliberate exceptions) is what would stop the next namespace going dormant unnoticed. That is the broader change and should be its own card if wanted.

Related: #4118 / PR #4137 (where this surfaced — the last live reader), #3810 and #4117 (dead fallbacks at live call sites — the mirror direction), #3546 (the ten-pack key discipline).

Generated by Claude Code