iMessage CloudKit per-thread sync stall: daemon restarts insufficient, Sync Now required

[BaseInfinity]

Cross-filed from BaseInfinity/claude-rdlc-wizard#3 — same fix applies to Codex-side RDLC projects.

Problem

Per-contact CloudKit sync stalls silently when ingesting iMessage from macOS chat.db. Some contacts current, others lag hours/days behind — no error signal.

What works

Messages > Settings > iMessage > Sync Now — Apple's official force-pull. Daemon restarts alone don't reset per-zone sync tokens.

4-step escalation ladder

1. WAL checkpoint: `sqlite3 ~/Library/Messages/chat.db 'PRAGMA wal_checkpoint(PASSIVE);'`
2. Kill imagent + identityservicesd
3. Kill full cascade: `killall IMDPersistenceAgent soagent cloudd`
4. AppleScript Sync Now button click (see claude-rdlc-wizard#3 for full script)

Smoke test fix

`imessage_account_sync_smoke.py` grouped by `h.account` — column removed in modern macOS. Fix: `GROUP BY h.id` (phone/email) for per-contact granularity.

Full script and details: BaseInfinity/claude-rdlc-wizard#3