[devin-ai-integration]

What and why

docs/project/performance-release-0.12.0-vs-0.11.0.md named develop e022cbca0 as the 0.12.0 candidate (#1788). Since then develop took #1793, #1794, #1792, #1788, #1795, #1791, #1787, #1775, #1752, #1778, #1797 and #1785, six of them Go on measured paths. This amends the record's "After the measured heads" section to name develop d16ca71e0 as the candidate and account for each of those merges, with the SHA each row set was measured at:

• fix(deps): resolve minimatch 11.0.1 in the VS Code extension and the Node client #1793, fix(deps): move the AsciiDoc converter to Asciidoctor.js 4.1.1 #1794, refactor(vscode): schedule the mass table's roll-up levels on one timeline #1787, docs(perf): account for every merge up to develop e022cbca0 in the 0.12.0 record #1788, chore(deps): keep @testing-library/react out of the grouped npm update #1752, chore(docs): revert the sponsor links to OpenMBEE's Open Collective page #1775 — Go-neutral by diff (git diff --stat … -- '*.go' go.mod go.sum quoted).
• fix(export): an expression body's parameter states its derived isVariable and mayTimeVary #1792 (export) — six-round re-probe at 86fa09ec9: neutral; WriteAPIJSON output byte-identical, ConvertAPIJSON/32 and /128 bytes and allocations unchanged.
• fix(export): the usages a declaration implies state isVariable/mayTimeVary from their owning type; a requirement's subject derives Usage::mayTimeVary #1795 (semantics + export) and fix(runtime): a false assumption makes a requirement hold (assumptions imply constraints) #1791 (runtime) — every row set re-run at a37809a6c against 86fa09ec9: load, REPL, derived and runtime rows unchanged in bytes and allocations (LoadModel/1000 563.8 k allocations on both, so fix(semantics): Usage::mayTimeVary derives over the effective isComposite — a performed action or exhibited state varies in time #1784's finding stands as reported); BatchSatisfy gains one allocation per assertion (+0.03 %, fix(runtime): a false assumption makes a requirement hold (assumptions imply constraints) #1791's verdict); the export rows grow +0.06–0.15 % (WriteAPIJSON) and +0.11–0.12 % (ConvertAPIJSON), the cost of fix(export): the usages a declaration implies state isVariable/mayTimeVary from their owning type; a requirement's subject derives Usage::mayTimeVary #1795's mayTimeVary derivation per implied usage.
• feat(grpc): trace and events for ExecuteAction #1778 (runtime + gRPC) — six rounds of the runtime and gRPC packages and the tests/perf gRPC rows at df2bc173c against a37809a6c: bytes and allocations equal or ~ on every row; binary +32 KiB.
• fix(semantics): a variant usage has no owning type, so Usage::mayTimeVary derives false #1797 (semantics, the IsVariant guard in UsageMayTimeVary*) and refactor(export): split writtenRelationships into per-source helpers #1785 (export, the element_ids.go split) — the load/derived rows and the export rows re-run at d16ca71e0 against df2bc173c: bytes and allocations equal or ~ on every load, derived, runtime and gRPC row (LoadModel/1000 563.8 k on both); WriteAPIJSON output byte-identical on four fixtures and 8 bytes larger on SimpleVehicleModel, which the diff of the two outputs attributes to fix(semantics): a variant usage has no owning type, so Usage::mayTimeVary derives false #1797 (its four variant usages' isVariable/mayTimeVary now false), with 40–50 fewer allocations per op on those two rows; ConvertAPIJSON/32 and /128 unchanged; refactor(export): split writtenRelationships into per-source helpers #1785 leaves no trace in any row.

Whole-binary floor, generated-model and Apollo 11 loads, the element-ID query and -convert ttl were re-run at each head (three runs after a warm-up) and stay inside their earlier spreads; sec/op rows that cross p < 0.05 do so on one column's spread and are not read, as the record's method says.

How it was verified

Docs only. python3 scripts/check-doc-links.py reports only the pre-existing third-party-notices.md notice link; no changelog fragment (docs/project record, per the previous two record PRs). Benchmarks: benchstat over six interleaved rounds per comparison with hypervisor steal logged per round (0–3.7 %), the stress ConvertAPIJSON/satellites=512 skipped as in the earlier probes.

Checklist

• make test and make lint pass locally (docs only; link check run)
• Tests added or updated for the change — n/a
• Documentation extended where it already covers the surface
• Changelog entry — not needed for a docs/project record
• baselines regenerated — n/a
• No internal work-item labels in the body, docs, or changelog

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.