Add non-convergence telemetry sink to council_verdict.js ESCALATED path (D.v2)

filter rejected GATE reversible: simple 3h proposed 21 May 2026

What is the proposed change?

At the existing ESCALATED verdict path (lines 386-397 per code inspection), after setting verdict to ESCALATED, add a file write to meta_engine/data/non_convergent/{hypothesis_id}_{timestamp}.json. JSON payload must contain: (1) full 3-round debate transcript, (2) per-axis score deltas between the most-disagreeing model pair (to identify which axis drove non-convergence), (3) hypothesis text and ID, (4) any partial kill rationale from rounds that reached consensus. No DB schema migration required — filesystem write only. Create the directory if it does not exist. Wrap in try/catch; log warning on failure but do not throw (must not block the ESCALATED verdict itself).

Target files

hypothesis_engine/moves/council_verdict.js

Expected effect

After 30 days, meta_engine/data/non_convergent/ will contain a corpus of ESCALATED transcripts. If the axis-delta field shows the same 1-2 axes driving disagreement across ≥60% of cases, those axes have ambiguous rubrics that can be sharpened. If zero files appear despite ESCALATED verdicts in logs, the sink is silently failing.

Falsifier — what would prove this wrong?

After 30 days: if council_verdict logs show ≥2 ESCALATED verdicts but non_convergent/ directory is empty or missing those files, the write is silently failing. If all transcripts show non-convergence concentrated on one axis in >80% of cases, that axis needs rubric clarification — the telemetry has done its job and the finding is actionable.

Evidence that triggered the proposal

Corpus D: brain/META_ENGINE_S158_ROUND2_SYNTHESIS.md — D.v2 COMMIT_WITH_REVISION: non-convergence telemetry explicitly approved but listed as not yet built
Corpus D: brain/META_ENGINE_S158_RED_TEAM_SYNTHESIS.md — net survivors from Round 1 include non-convergence telemetry (from Gemini judge); Round 2 elevated to COMMIT_WITH_REVISION
Corpus E: council verdicts last 7d — 5d7cca deferred verdict indicates non-convergence cases are occurring but leaving no analysis trail for meta-engine learning

Proposer self-score

The proposer scored its own draft on these axes (0-3 each) before submitting.

Axis	Score
specificity	3
falsifier	2
solo feasible	3
blast radius	3
composability	3
reversibility	3

Disposition

Rejected by filter_score. The proposal did not meet the bar for specificity, falsifiability, or solo-feasibility.

Evaluation history

When	Move
2026-05-21 04:16	meta_filter_score
2026-05-21 04:15	meta_genesis