docs(rfc): amend cost harness with the DST benchmark

[azimafroozeh]

What & why

This PR amends RFC-031 in place: a new §11 records the counting side as built, and nothing in §§1-10 changes.

• Two concurrent implementations of the RFC's counting idea now exist: test(bench): add node-vs-edge logical cost v0 #503 on real backends, and the DST benchmark's counting pass, hosted on the harness proposed in docs(rfc): add RFC 0037 deterministic simulation harness #507.
• §11 records the first per-operation cost profile, the calls-only golden decision, and the first measurement placing both instruments on one ruler (writes match exactly; one named residual stays open).
• It states the division of roles (per-PR logical comparator versus real-backend truth, converging at the benchmark's v2 with test(bench): add node-vs-edge logical cost v0 #503's techniques credited) and why the golden is a structural gate, not a performance gate.

Local verification

• Markdown-only diff to docs/rfcs/0031-comparative-cost-harness.md; cargo gates not run: no code touched
• bash scripts/check-agents-md.sh: 77 links, 72 docs OK

Greptile Summary

RFC-031 is amended with implementation evidence and a proposed division of responsibilities between deterministic logical-call counting and real-backend cost measurement.

• Documents the DST and PR test(bench): add node-vs-edge logical cost v0 #503 counting approaches and their initial measurements.
• Records a calls-only golden decision and outstanding reconciliation differences.
• Describes convergence toward provider-specific gates while retaining release-profile timing.

Confidence Score: 5/5

The documentation-only PR appears safe to merge.

The amendment scopes the calls-only golden to the deterministic instrument, preserves the broader RFC requirements, and clearly identifies referenced harness work as landing through an in-review PR.

Important Files Changed

Filename	Overview
docs/rfcs/0031-comparative-cost-harness.md	Adds a documentation-only amendment describing completed counting work, benchmark observations, instrument responsibilities, and CI posture without introducing an actionable defect.

_{Reviews (1): Last reviewed commit: "imp" | Re-trigger Greptile}

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.