Fix Journal composer limits, blank drafts and recovery guidance

[RCmerci]

The shared Journal composer left attachment actions looking active at nine items, admitted Unicode-only blank drafts, labeled the multiline Return key as Send, and offered reselection even when a failed batch could be retried in place.

This change disables Files/Photos/Camera at the cap and explains how to restore them, uses Unicode White_Space consistently for admission and attachment-only naming, keeps native Return for multiline editing, and chooses recovery guidance from the retained import state. Worker interruptions preserve the correct recovery instruction; nonstaged failures retain reselection guidance.

Current integration: head 601a114f27b0e39b66b01aaf6f2b355fea70c5a7 incorporates main c1276ec28ea0a28804db7b8bcd1df0b1cb725915. The sole conflict was appended route tests; both suites and registrations are preserved. The production fix is equivalent to the original PR64 patch. PR63's typography, Timeline indicator, ten-level Detail expansion and visible-demand read budgets are preserved.

Validation:

• RED public reducer/view tests and four actual baseline Simulator reproductions before implementation.
• Current head: dune build @all app/native_embed.exe.o and complete dune runtest test --force -j 2 passed, including 41 routes and 79 Application tests and the PR63 read-budget assertions.
• Current-source iPhone 13 / iOS 26.1 Simulator: all 8 affected acceptance tests passed, zero failures, 342.930 seconds. Coverage includes all three entries for cap/remove/re-add and native Return, Unicode blank admission, text and attachment submissions, duplicates/removal, Detail back targets, recoverable import failure and Worker interruption.
• Independent incremental review verified the remote SHA/base, 68 source hashes, host binary, both test suites and all 8 raw xcresult leaf results; no evidenced code defect or code blocker.
• Historical original head 5425b1e: the broader 16-test Simulator matrix passed before integration. That run is retained as historical evidence, not attributed to the current head.
• Exact-head CI: Journal CI run 38138798348 completed/success; WIP and build-and-test both passed for the exact head.

Tests use an isolated synthetic graph/account and dedicated app bundle with the actual Application → C → Swift/LUI host; native keyboard and pickers are driven through XCTest. The nonstaged reselection branch is covered through its public callback/Worker completion, not a visible iOS entry point.

Limits: IME composition, VoiceOver, physical iPhone, live server/sync, successful camera capture, actual imported bytes and final asset persistence are untested. Photos tests used only synthetic QA images. Dependency pins, Dune and spec files are unchanged.

[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 2026-10-11T09:28:46.309043Z	5425b1e	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.