Phase 23.1: Clean up evidence pipeline contract

[jununfly]

This was generated by AI during triage.

Parent

• Phase 23: Optimization architecture cleanup #218
• Post-PRD optimization tracker for Rust indexing performance #165

What to build

Tighten the benchmark/evidence tooling contract introduced in Phase 22 so future Rust indexing optimization candidates can reuse artifact comparison, ranking, and decision-draft output without hand-written interpretation.

The toolchain must remain local-only. It must not call GitHub, require network access, or update trackers automatically.

Acceptance criteria

• Consolidate duplicated artifact parsing and comparison helpers introduced by the evidence pipeline.
• Normalize target status, required/stress classification, empty-corpus status, sufficiency, graphStats parity, wall time, RSS, and unavailable-reason output.
• Keep rollout-readiness disclaimers explicit in generated output.
• Keep comparison, ranking, and decision-draft output reusable for later Post-PRD optimization tracker for Rust indexing performance #165 candidates.
• Do not add GitHub/network behavior to the tooling.
• Do not add new benchmark dimensions unless they directly support candidate selection.
• Add or update fixture tests for artifact parsing and comparison output.
• Add or update fixture tests for ranking and decision draft output.
• Smoke-generate output using existing Phase 22 artifacts.
• Do not run a full VS Code sparse scoreboard for this slice.

Blocked by

None - can start immediately.

[jununfly]

This was generated by AI during triage.

Completed as part of Phase 23.

Summary:

• Evidence output now includes an explicit local-only Evidence Contract section.
• RSS unavailable reasons are now surfaced in generated comparison output.
• Performance experiment paths are inventoried as production path, retained diagnostic, or dead candidate.
• Production cleanup stayed on the benchmark/evidence boundary and did not change default indexing behavior.
• Targeted smoke/profile artifacts were generated from existing Phase 22 evidence; no full VS Code sparse scoreboard was run.
• Profile Rust parseExtractionMs before next optimization #224 was created as the next Post-PRD optimization tracker for Rust indexing performance #165 step for parseExtractionMs profiling.

Artifacts:

• docs/benchmarks/2026-06-18-rust-indexing-core-phase-23-experiment-inventory.md
• docs/benchmarks/2026-06-18-rust-indexing-core-phase-23-targeted-evidence-smoke-comparison.md
• docs/benchmarks/2026-06-18-rust-indexing-core-phase-23-targeted-evidence-smoke-decision-draft.md
• docs/benchmarks/2026-06-18-rust-indexing-core-phase-23-closeout-decision.md

Validation:

• npx vitest run __tests__/rust-indexing-evidence.test.ts
• npx vitest run __tests__/rust-indexing-evidence.test.ts __tests__/rust-indexing-experiment.test.ts
• npm run build

#165 remains open. Rust default rollout readiness is not claimed.