100 metas across 24 tag files are mapped in properties and emitted into children, where createComponentRenderable filters them straight back out (packages/runes/src/lib/component.ts:127-131). 133 other metas already omit that emission and render identically.
Delete the emission. Nothing else.
Tracking started Sep 24 — check back for trends.
97231beproperties is also listed in a children array or pushed onto one, in any of the 121 runesrefrakt contracts --check reports no drift, on both committed copies (contracts/structures.json and packages/lumina/contracts/structures.json)npm run seo:baseline:check reports no driftnpm test passes unchangedproperties and is deliberately rendered is left alone — the sweep is bounded to property-mapped metasThe emission takes two forms and both must go: direct entries in a children: [...] literal (63 sites) and children.push(meta) / multi-argument pushes (37 sites).
Verified before filing by stripping all twelve from plugins/plan/src/tags/work.ts: refrakt inspect work --site plan --json returned byte-identical output and npm run seo:baseline:check reported no drift.
The deletion is safe by construction, not just by sample. schemaTags (component.ts:57) is populated only from result.schema, which has zero callers — so isSeoMeta is always false, every property meta is added to pureDataMetas, and pureDataMetas members are unconditionally filtered from childArray. There is no input for which a property-mapped meta survives.
Highest-count files: work.ts (12), bug.ts (10), map.ts (8), decision.ts (8), spec.ts (8), form.ts (6).
contracts --check plus seo:baseline:check returning no diff is the test. Both artefacts are generated from config and describe every rune's complete output, so a behavioural change cannot hide from them.
data-rune-fields bag that made the emission inert