⚡ #543 remeasure weights and performance, then deliver

[taras]

Parent quest: #543

Purpose

Integrate the completed PC, CT, and IE migrations, prove the frozen matrix, measure the result under the current runtime topology, and deliver #543.

Matrix audit

• Account for every PC, CT, IE, and DT row exactly once.
• Confirm all DT rows remain focused TypeScript command tests pending Add in-process xmd command execution to Markdown tests #581.
• Confirm no row in these four suites is labeled a permanent OS-process claim.
• Confirm launchers contain no row assertions and TestResult is unchanged.

Focused feedback

Run the three launchers, the four retained suites, and testing-execution-host.test.ts under Deno; run the three launchers under Node and Bun. Create the feedback commit as soon as this frozen evidence and all required mutation probes pass.

Performance evidence

• Repeat three alternating local /usr/bin/time -p deno task test samples on clean before/after worktrees and compare medians.
• From exact-commit hosted runs, record every Deno shard Test window, their sum, and the slowest shard before/after.
• Slowest-shard change is an observation, not a gate.

Weights and partition

• Dispatch the hosted weight-measurement workflow against the implementation ref.
• Commit the generated test-weights.json artifact byte-for-byte with full provenance.
• Recompute Deno, Node, and Bun partitions; record file counts, serial totals, affected file weights, fallback set, and every predicted shard total.
• Prove pairwise disjointness and exact applicable-corpus union with the existing discovery/weight/shard/workflow regressions.
• Apply the specification's five-run fixed-head calibration rule if any installed runtime count no longer qualifies.

Acceptance

• Frozen matrix, focused evidence, mutation records, local timing, hosted Deno spend/slowest shard, generated weights, and partition evidence are durable in the PR or linked issues.
• Required CI is green under the authoritative workflow.
• 🗺️ Quest: Prove document behavior through <Execution>, and spawn xmd only for process claims #543 can close without waiting for Add in-process xmd command execution to Markdown tests #581.

Dependency

Blocked by #584, #585, and #586. Final task in the quest.

[taras]

Delivered: PR #606 merged at e425ee4 (feedback commit 73b73ed). Fresh test-weights.json (run 32873071843 at 170a021, byte-for-byte), shard counts calibrated to 9/5/2 by the five-consecutive-run rule (7/3 and 8/4 missed at the committed weights; run 32892093466 attempts 1–5 all pass), final hosted after-measurement recorded against the #582 baseline (slowest Deno shard 330s → 270s), final partitions disjoint with exact union and an empty fallback set. Matrix audit, focused three-runtime evidence, and alternating local medians are in the PR body and issue-587-implementor-handoff.md.

[taras]

Planner closure — PASS

#587 is complete on main at merge commit e425ee42ebc024ff7a092d0fcbbe769436a346ee (PR #606; reviewed head 73b73edef0b3090928410b6ea0ea9af7512f0622).

• The frozen PC, CT, IE, and DT disposition matrix accounts for every row exactly once; DT remains focused TypeScript coverage pending Add in-process xmd command execution to Markdown tests #581, and no row is classified as a permanent OS-process claim.
• The three launchers contain only suite-level success, non-empty-result, and all-passing assertions; the public testing surface and TestResult are unchanged.
• The alternating local whole-suite record compares three green samples on each side. Aggregate medians were 1598.84s before and 1612.37s after within an approximately 90-second sample span; the directly affected suites fell from approximately 110s to 94.2s despite concurrent corpus growth.
• The final hosted Deno comparison records every Test window: 2112s before and 2214s after, with the slowest shard improving from 330s to 270s as an observation rather than a gate.
• test-weights.json is byte-identical to hosted measurement artifact run 32873071843, with an empty fallback set.
• Mechanical fixed-head calibration rejected 7/3/2 and 8/4/2, then proved 9/5/2 as the smallest qualifying Deno/Node/Bun counts through five consecutive attempts of run 32892093466.
• Final partitions are pairwise disjoint and exactly cover 327 Deno files and 254 Node/Bun files; all eight focused partition suites passed.

#587 frozen acceptance and delivery evidence are satisfied.