** This file was created with an agent. **

#Handover — Phase 4 of the inline-node refinements plan: the viewer channel

Status: Sub-RPI handover. Written 2026-10-13 for seeding a fresh /research_codebase → /create_plan sub-cycle on the remaining viewer-attach failure. Everything below is verified state as of HEAD 0d012dd4 (branch feature/composition-and-justification).

#The task in one paragraph

Approach (a) of the inline-node refinements: inline nodes (cite, bdi, generic-inline, reproducing atoms) must style like marks — through a third style-link family inlineNodesStyleLinks resolved via the styleLinkProperties@ merge channel and applied by UIDocumentStyleStyler — instead of the per-node block styler (UIDocumentTypeSpecStyler), whose block-layout writes (width/text-align/block margins/root-fallback font-size) are wrong for them. Phases 1, 2, 2b, 3 are COMMITTED and green; phase 4 is implemented in the working tree and ~90% verified: the only remaining failure is a fresh viewer attach after a mode switch (live edge edits in the viewer all pass; editor passes).

#State as of this handover

#Committed (green; do not redo)

#Uncommitted working tree (reconstructed CLEAN, all green)

#Verified green

• npm run typecheck 0 errors; npm run lint clean; vitest 55 files / 541 tests (incl. 2 new third-family tests, 7 pair tests, the merge-channel em gate test).
• Probe qa-inline-node-style-links.mjs: 5 of 6 steps pass — baseline a1 (no block-writes on inline), live edge add (all 153 inline elements style at the patch’s 30pt), live tombstone (clears), live re-link (restores), editor mode with the stable edge (30pt), no PAGEERROR.

#The one remaining failure

back to viewer (stable edge): after switching editor→viewer with a stable inlineNodesStyleLinks edge, all 153 [data-node-type="generic-inline"] elements render unstyled (styleAttr: null, styled: 0). The styling machinery is fine everywhere else — this is a fresh-attach provisioning gap: the fresh UIDocumentElement attachments’ _provisionInlineStyleStyler must resolve the edge (its SLM registration exists — VWRES-CALL resolutions log non-null ids, 765× during one instrumented run), yet the styler never applies.

#Mechanics discovered (the hard part — read before researching further)

1. The channel itself works: SLM registration (StyleLinksMeta, third instantiation) + merged payloads (StyleLinkLiveProperties → createPatched against the containing typeSpec) + UIDocumentStyleStyler write path are all proven by the passing steps.
2. The wake gap + its fix: viewer elements created at boot never re-provisioned on a live edge add. Root cause: _syncMetaInfo’s re-provision was gated on an existing wrapper (oldWrapper !== null). Fix applied (and REQUIRED — steps 2–4 fail without it): unconditional inline re-provision on pathChanged || scopeSettled. The meta tree’s _scopeSettled() (document-nodes-meta/index.mjs:839) refreshes all attachments (refresh(handlerFn, metaInfo, true)) when the node’s own node-properties scope settles — that refresh drives the re-provision.
3. The rebuild-compare identity: the styler wrapper’s identity is the resolved styleLinkProperties@ id; rebuild on id change; the containing-typeSpec GenericUpdater trigger mirrors the composition@ pattern.
4. Caveat discovered (console-log archaeology is dangerous): an intermediate debugging phase left a composition regression (14 failing tests) that bisected to a damaged viewer.typeroof.jsx; the current file is a CLEAN reconstruction from HEAD + the deliberate changes above and the battery is green. The composition suite (text-composition) is the canary — it must stay green.
5. Residual open questions for the sub-cycle:
   • Why does the fresh attach (mode switch) not provision the inline styler while live-edit scopeSettled re-provisions work? Is _provisionWidgets on fresh attachments not reached for inline elements (instrumented VWPROV call-count was 0 once), or does the fresh attach provision BEFORE the typeSpec tree settles (the widget order: TypeSpecMeta → DocumentNodesMeta → renderer — so it should be settled), or is the wrapper built but its initial update lacks properties@ (gate in UIDocumentStyleStyler.update needs properties@ or verboseFontVariationSettings in the changedMap)?
   • At fresh attach: is hasRegistered(styleLinkProperties@…) false at provision time while true later (SLM re-provision lag on mode switch)?
   • The TRIGGERS made=776 fired=0 observation: protocol setUpdated does NOT independently wake wrappers — it piggybacks on state cascades; check whether the trigger wrapper is pruned from cascades (its only dependency is a protocol id).

#The plan (upstream)

• thoughts/plans/2026-10-12-2127-inline-node-refinements-approach-a.md — six phases; phases 1, 2, 2b, 3 done and committed; phase 4 here; phases 5 (“Node Style Links” UI wiring) and 6 (composition basis threading for nested leaves + QA + state re-tuning) remain.
• docs/planning/inline-node-refinements/ROADMAP.md — the design conversation + all operator decisions (pair form, channel split, keep baseFontSize, no rename, unconditional a1 skip, the inlineNodesStyleLinks pivot, one-element inline DOM).
• thoughts/research/2026-10-12-1723-font-size-resolution-and-rebasing.md — the research round (font-size synthesis, nodeProperties channel, em/baseEm/en resolution sites, inline/patch basis, migration inventory).

#Suggested sub-cycle seed

/research_codebase scoped to: viewer attachment lifecycle around mode switches — where fresh UIDocumentElement attachments provision, when _syncMetaInfo/_scopeSettled fire relative to StyleLinksMeta registration, and whether the trigger wrapper is cascade-pruned (its protocol-only dependency). The canary probe is scripts/puppeteer-probes/qa-inline-node-style-links.mjs — make the “back to viewer (stable edge)” step pass; keep everything else green (typecheck/lint/541 vitest + the 5 passing probe steps).