Add genesis backend-fallback harness for context overflow

filter rejected HARNESS reversible: simple 5h proposed 31 May 2026

What is the proposed change?

Wrap the genesis LLM call in a fallback chain: try configured primary (currently oauth_opus, 200K ctx), on context-overflow or 4xx-context-related error retry with bedrock_opus_46_1m, on second failure retry with bedrock_sonnet_46. Emit fallback_chain_used to engine.db move_log. Do not retry on transient 5xx (existing retry harness handles those). Add a counter row to engine_metrics so fallback rate is queryable.

Target files

hypothesis_engine/moves/genesis.js hypothesis_engine/lib/backend.js

Expected effect

Eliminates the silent 5-cycle genesis outage class observed in S180. Over 30 cycles, ≥98% of genesis attempts produce output. Fallback chain is exercised on <10% of cycles (otherwise input corpus is too large and needs trimming, not fallback).

Falsifier — what would prove this wrong?

Track fallback_chain_used over 30 cycles. If fallback never fires, the harness is dead weight and the primary backend alone is sufficient — remove. If fallback fires on >25% of cycles, the primary is mis-sized and the fix is corpus trimming, not fallback.

Evidence that triggered the proposal

D — commit 0f2d20d 'S180: fix meta_engine genesis 5-cycle silent outage'
D — commit ca950ff 'S180: genesis runs free on OAuth Opus — config-selectable backend'
E — S180 5-cycle outage in move_log

Proposer self-score

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

Axis	Score
specificity	3
falsifier	3
solo feasible	3
blast radius	2
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
2026-05-31 04:05	meta_filter_score
2026-05-31 04:03	meta_genesis