Executed evidence
zig build benchmark-comparison-test --summary all reached the repository-owned validators after the Home runner fix, then failed closed because bench/comparison_zig_js.zig no longer matched representative matrix V46. Commit a4d47ae added a sparse runtime_operations map to each tier-attribution row while leaving the collector checkpoint at schema 17 and without validating or differencing that map. The aggregate execution.runtime_operation_calls was validated, but the newly published per-opcode structure was accepted as an unversioned, unchecked field.
Scope
- Publish representative matrix V47 with the updated shared native-runner pin and no scored workload, checksum, job, timing, or execution-boundary change.
- Add a nested schema-v1 diagnostic contract for
runtime_operationswhile retaining sidecar schema 17. - Model
runtime_operationsexplicitly in snapshots and phase deltas. - Fail closed on non-integral, negative, regressing, or aggregate-incoherent per-opcode counts.
- Preserve the frontend runner pin and every unrelated integration contract.
Acceptance
- Matrix structural tests reject runner, diagnostic-schema, and per-opcode attribution drift.
- Attribution self-tests cover invalid values, aggregate mismatch, and counter regression.
zig build benchmark-comparison-test --summary allpasses all repository benchmark validators.- Benchmark documentation names the current frozen matrix and diagnostic boundary accurately.
No test262 behavior is changed.