docs(perf): measure 0.10.0 against 0.9.2

[devin-ai-integration]

What and why

The release-gate performance measurement of the 0.10.0 candidate against release 0.9.2. The candidate is hotfix/0.10.0-develop-sync at 82f0fac87 (#1024): main at the 0.10.0 release merge with develop's post-cut changes and the six performance fixes (#1017–#1023) folded in. The baseline is the v0.9.2 tag, built and benchmarked from its own worktree. main at ffe51571c — the release merge alone — was measured as a third revision and is shown as a column wherever a fix moved a row, so the record says what each fix bought.

Method as in the 0.9.2 against 0.9.1 record: every benchmark package on every revision under benchstat over six counts, every row in doubt re-run interleaved across the three, the stress constellation and tests/perf workloads at the previous record's sizes, and whole-binary sysml -validate wall time and RSS on generated models, the examples and the Apollo 11 model, start-up and binary size. A benchmark that fails on one revision prints no result line and the package's other rows still run; such rows are reported as candidate-only or baseline-only, nothing was excluded from -bench ..

Verdict: nothing here blocks the release. Nothing scales worse than it did; the whole binary validates a 12 000-declaration model in 0.28 s against 0.23 s. The six fixes cleared three of the regressions the release merge alone showed and cut a fourth by a quarter: CompiledCalc/Fib(25)/interpreted from 5.3× 0.9.2 to 20% faster than 0.9.2, BatchConstraints from +68% to ~, GRPCParseFileCached from 299 allocations back to 24 (0.9.2: 22), the workspace-edit rows from 3.8–3.9× 0.9.2's bytes to 2.9–3.0×. What remains, confirmed interleaved and attributed in the record:

• loading +15–18% allocations and +10–16% time per element on connector-dense models (connector ends owned, journaled, indexed and reflected) — structural, linear;
• executing an action or state step +20–60% with 1.3–1.5× the allocations (statement-order exploration, body-binding frames, held control values) — the release's semantics; the per-step constant is the next place to look;
• editing beside a large document still 2.9–3.0× the bytes (the per-name snapshot of re-export claims) — the one row with an evident fix still open;
• the import-ambiguity warning fires 134 times on the Annex A vehicle example; Analyze/vehicle +85%, ConnectParseFileHTTPCached 271 → 879 µs (no JSON cache on the HTTP path);
• exact numerics: Collatz(27) +59–65%, Hypot/go 11 → 56 ns, Instantiate +11–31% per object;
• for the release notes: SumTo(1000000) no longer fits the default step budget, the C target refuses Integer calcs, and the stress benchmark's per-document analysis trips the new decimal-rounding lint.

No regression is fixed here; this PR is the record only. It is linked from docs/internals/performance.md, docs/project/README.md and the mkdocs nav.

How it was verified

python3 scripts/check-doc-links.py (0 broken links), make docs-check, python3 scripts/changelog.py check. The measurements themselves: Xeon Platinum 8559C, 8 CPUs, 31 GiB, Go 1.25.11, GOMAXPROCS=8, the revisions never run concurrently; the tables are the benchstat rows with their p-values.

Checklist

• make test and make lint pass locally (docs-only change; make docs-check passes)
• Tests added or updated for the change (not applicable)
• Documentation extended where it already covers the surface (see CONTRIBUTING.md)
• Changelog entry added as changes/unreleased/<slug>.<section>.md, not as an edit to CHANGELOG.md
• baselines regenerated and make docs-counts run if a gate count moved (none moved)
• No internal work-item labels (waves, slices, F4, K5) in the body, docs, or changelog

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/60e111c2f4c043b887f52d4aeadc556e
Open in Devin Desktop: https://nasa-jpl-demo.devinenterprise.com/desktop/session/60e111c2f4c043b887f52d4aeadc556e?variant=devin
Requested by: @HuiJun

[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: 3 flags

Not posted on this PR by your GitHub settings — view them in Devin Review. (Configure)