Post-PRD optimization tracker for Rust indexing performance

[jununfly]

This was generated by AI during triage.

What to build

Create the Phase 15 tracking plan for controlled performance gate interpretation after #161 rerun5. This phase preserves the existing Phase 14 A/B experiment structure and raw PRD performance gate, but adds a Rust-side Graph Work Profile control so matched-work evidence can distinguish implementation performance from extra graph work.

This tracking issue follows the Rust Indexing Core Vertical Slice PRD and #161 rerun5 outcome. #161 remains open until required-target trigger conditions are met or explicitly redefined by the maintainer.

Acceptance criteria

• Track graph work profile plumbing from manifest through CLI into Rust core.
• Track matched-work required-only experiment evidence for zcodegraph and excalidraw.
• Track controlled interpretation of raw full-profile evidence vs matched-work evidence.
• Keep VS Code stress validation deferred until Phase 14: Run required targets and one VS Code stress validation #161 trigger conditions are met or explicitly redefined.
• Do not close or modify Phase 14: Run required targets and one VS Code stress validation #161 as part of creating this plan.

Blocked by

None - can start immediately.

[jununfly]

This was generated by AI during triage.

Phase 15E has been split and published for RSS gate closure work.

Plan: docs/plans/2026-06-16-rust-indexing-core-phase-15e-rss-gate.md

Implementation issues:

• Phase 15E.1: dhat-rs heap profiler integration #170 — Phase 15E.1: dhat-rs heap profiler integration
• Phase 15E.3: dhat-heap.json to HTML summary report #171 — Phase 15E.3: dhat-heap.json to HTML summary report
• Phase 15E.2: SQLite batching + WAL pragma for Rust indexer #172 — Phase 15E.2: SQLite batching + WAL pragma for Rust indexer

Tracker:

• Phase 15E: RSS gate closure tracker #173 — Phase 15E: RSS gate closure tracker

Important guardrail: Phase 15E is diagnostic + bounded optimization work for the RSS gate. It must not claim default or full-profile rollout readiness. #165 should remain open until the RSS gate is actually closed or the Rust indexing architecture is explicitly reassessed with evidence.

[jununfly]

This was generated by AI during triage.

Phase 15E has local implementation + rerun4 evidence ready for review.

Key artifacts:

• Plan decision: docs/plans/2026-06-16-rust-indexing-core-phase-15e-rss-gate.md
• Rerun4 raw artifact: docs/benchmarks/2026-06-16-rust-indexing-core-phase-15e-vscode-matched-work-stress-rerun4.raw.json
• Rerun4 summary: docs/benchmarks/2026-06-16-rust-indexing-core-phase-15e-vscode-matched-work-stress-rerun4-decision-summary-draft.md
• Heap report: docs/benchmarks/phase-15e-rerun4/dhat-heap.json
• Heap HTML summary: docs/benchmarks/phase-15e-rerun4/dhat-summary.html

Result:

• VS Code matched-work stress sufficiency passed.
• RSS trend improved from Phase 15D +20.08% Rust-over-TypeScript to Phase 15E -21.41%.
• The PRD RSS gate is still not closed because Rust is not yet at least 30% lower than TypeScript.
• Wall time in this rerun includes dhat heap profiling overhead and is not a rollout-readiness wall-time signal.
• Phase 15E classification is continue-with-batching-only.

Do not close #165 yet, and do not claim default or full-profile rollout readiness from this phase.