fix(storage): keep equivalents quarantined-ambiguous when no head is accepted

[Sinity]

Summary

When classify_membership_revisions refuses to pick a head (accepted_raw_ids=()), the membership write-back no longer labels equivalent_raw_ids as superseded_equivalent/byte_proven — they collapse back into the unresolved cohort as ambiguous/quarantined.

Problem

Live-archive forensics (Ref polylogue-yla8; report .agent/reports/byte-headless-914-investigation-2026-07-21.md) found 914 logical sources (155,816 max messages) whose memberships read byte_proven/superseded_equivalent yet have no raw_revision_heads row, no application receipt, and no session — supersession authority fabricated for a head that was never written. Introduced by 6a579d0 (#2684, 2026-07-11); the 2026-07-20 rebuild walk amplified it (93% of the cohort decided in one hour).

Solution

One-line gate in apply_raw_membership_classification's decisions construction: equivalents map to superseded_equivalent only when accepted_raw_ids is non-empty. The authority ternary then keeps them quarantined automatically. Comment documents the incident.

Verification

• devtools test tests/unit/storage/test_revision_replay.py → 23 passed.
• New test_headless_cohort_keeps_equivalents_quarantined_ambiguous: all-ambiguous cohort with duplicate content → no head row, every member ('ambiguous','quarantined'). Anti-vacuity: with the fix stashed, the test fails (pre-fix labeling) — verified.
• devtools test tests/unit/storage/test_no_string_interpolated_sql.py (archive.py self-audit) → pass.
• devtools verify --quick → exit 0.

Note: repairing the existing mislabeled rows on the live archive is a separate coordinator-owned step (re-census after this merges).

Ref polylogue-yla8

🤖 Generated with Claude Code

https://claude.ai/code/session_01QUsH3Rhq6oAZpYPWcsZqnZ

Summary by CodeRabbit

• Bug Fixes

  • Improved handling of conflicting revision branches when no accepted version exists.
  • Keeps affected records marked as ambiguous and quarantined instead of incorrectly treating them as superseded.

• Tests

  • Added coverage to verify that headless revision groups retain their ambiguous status and do not produce an accepted revision.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: f933e23e-b00e-405c-aa9c-51b8973ed94c

📥 Commits

Reviewing files that changed from the base of the PR and between 2f3aa34 and f5e31cb.

📒 Files selected for processing (2)

• polylogue/storage/sqlite/archive_tiers/archive.py
• tests/unit/storage/test_revision_replay.py

---

📝 Walkthrough

Walkthrough

The archive now labels equivalent raw memberships as ambiguous when classification produces no accepted raw head. A regression test verifies that headless cohorts persist without an accepted head and retain ambiguous, quarantined membership authority.

Changes

Headless membership classification

Layer / File(s)	Summary
Conditional equivalent decision and regression coverage polylogue/storage/sqlite/archive_tiers/archive.py, tests/unit/storage/test_revision_replay.py	Equivalent raws receive superseded_equivalent only when an accepted head exists; otherwise they remain ambiguous. The test verifies no accepted head is stored and memberships remain ambiguous with quarantined authority.

Estimated code review effort: 2 (Simple) | ~10 minutes

Possibly related PRs

• Sinity/polylogue#3165: Both modify the archive membership classification and replay path.
• Sinity/polylogue#3204: Both adjust membership-head and receiver decision labeling in archive.py.
• Sinity/polylogue#3234: Both address classification outcomes when no accepted raw head exists.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title is concise and accurately summarizes the PR’s main behavior change.
Description check	✅ Passed	The description covers Summary, Problem, Solution, and Verification clearly, with only non-critical sections omitted or folded into notes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/storage/headless-cohort-authority-mislabel

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f5e31cb0f5

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Count equivalent members in quarantine totals

When a headless cohort contains duplicate content, this branch now quarantines every equivalent_raw_id, but backfill_historical_revision_evidence still increments its quarantine total using only len(classification.ambiguous_raw_ids) at polylogue/sources/revision_backfill.py:863-864. For the new three-member fixture, all three membership rows are quarantined while the result reports only two, so raw_materialization_quarantined_count and the operator-facing repair detail understate the unresolved evidence; include the equivalent IDs in that count when no head is accepted.

Useful? React with 👍 / 👎.

[Sinity]

Coordinator review: read the full diff. One-line semantic gate at the decisions construction (equivalents map to superseded_equivalent only when an accepted head exists) + incident comment; regression test asserts no-head cohorts stay ('ambiguous','quarantined') for every member, anti-vacuity verified via stash (pre-fix labeling fails it). archive.py self-audit test run explicitly per repo gotcha. Live-archive re-resolution of the 914 mislabeled rows is named as a coordinator follow-up, not claimed. Merging on green quick-gate.