[RCmerci]

Summary

Register journal pages even when their initial root read is empty. Previously these pages were missing from the runtime registry, so a page-tree-only sync invalidation did not refresh the initial feed. An older scan still loading history could then publish empty Today over the synchronized content.

Registration makes that invalidation start a new initial feed, reusing the existing feed-session fencing to retire the old scan. No Capture, UI, protocol or public API changes.

Add one public-runtime regression: discover empty Today, hold its initial history continuation, synchronize three Today roots, then finish the old continuation before or after the refreshed feed. Both orderings must retain the synchronized roots without Capture.

Validation

• Red/green: the new regression failed with [] instead of the three synchronized UUIDs before the production fix; all 184 runtime cases pass after it.
• dune build @all app/native_embed.exe.o — passed.
• dune runtest --force — passed (1,088 Alcotest cases across 21 suites).
• These commands used the isolated prepared OCaml 5.5.0 switch and external build directory; repository dependency pins and Dune files are unchanged.
• spec-dev-tool check --all — 297 valid documents; 5 pre-existing invalid documents, each byte-identical to the base revision. No decision documents changed.
• Live Web/iOS Simulator comparison: the baseline reproduced Current + No journal entries followed by Capture recovery when existing blank roots received text. The fixed build automatically displayed the two edited roots in six cold-start runs without Capture. Public graph export later confirmed an additional blank root was retained; this was not a two-total-root graph.
• Three further fixed-build cold starts began with Today roots/members=0, verified through public Graph EDN export after native deletion and synchronization. After exiting the app, Web created three new roots each round. All three appeared automatically and remained stable for over 60 seconds after Current, without Capture. The 30 historical dates / 381 children were preserved.
• Simulator executable, native producer and import manifests were rechecked against source/native source d135571d46abbd4c3c9b146e61874f5986170347; installed executable SHA-256 15f2f51c2991a9ebfbbe20a50034f05c2f4fc56d159ff43c454614664c663778.
• GUI evidence uses an experimental host-restamped Simulator build (cross_abi_validated=false), not real-device/cross-ABI acceptance. Public current-entity exports establish logical zero roots, not native database byte-level erasure. GUI runs do not establish internal response ordering or exclude every possible timing failure; controlled runtime regression covers the two response orderings.

Link to Devin session: https://app.devin.ai/sessions/55da92d6550b4d348f51aff910621294
Open in Devin Desktop: https://app.devin.ai/desktop/session/55da92d6550b4d348f51aff910621294?variant=devin
Requested by: @RCmerci

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-28T06:06:39.311996Z	d135571	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.