fix(scale): reconcile measured lifecycle accounting and admission

[DecisionNerd]

Publication evidence previously omitted manifest installation and path-copy reads, while lifecycle admission assumed fixed CAS synchronization counts. This change carries measured payload, manifest and authentication components through the ordinary facade and reconciles their exact totals across both controlled 1×/2×/4× lifecycle axes.

Closes #1116
Refs #951

The implementation records actual file and directory barriers, preserves synchronized cache-window eviction, and accounts for a concurrent losing publisher's temporary work without counting reused objects as newly retained bytes. Retained shape allocation includes the runtime catalog. Checked arithmetic and protocol-derived Patricia bounds reject omitted, misclassified, duplicated, overflowed or excessive work; source/imported storage inventories remain authoritative.

Historical fixture readability is separate from authenticated certification. The parity gate checks the preserved lifecycle and complete phase mapping; certification still requires an externally anchored provider result. Regressions reject malformed/incomplete history and failed unmapped drills without fabricating provider evidence.

Validation on OVHC-AGENCY's ext4 filesystem:

• cargo test --locked -p graphforge-storage --lib -- --test-threads=4: 953 passed, two unchanged existing ignores; private mount namespace supplies ext4 /tmp for durable tests.
• cargo test --locked -p graphforge-api --lib -- --test-threads=4: all 681 passed, including exact payload/manifest/path-read reconciliation with unchanged growth ceilings.
• cargo test --locked -p graphforge-api --test scale_g500_ladder -- --test-threads=1 --nocapture: all 29 passed, including both 1×/2×/4× axes, publication, reopen, export, clean import, recovery and the million-edge sink.
• PYTHONPATH=benchmarks/harness uv run pytest -q scripts/ci/test-validate-g500-certification.py benchmarks/tests/test_scale_parity.py: 258 passed, plus 14 adversarial subtests.
• make -C benchmarks smoke-python: 389 passed.
• make -C benchmarks smoke-rust: all Rust benchmark tests, Clippy, formatting and the public-facade dependency boundary passed.
• Fresh CLI and all benchmark binaries built from this tree; benchmarks/scripts/test-tiny-lifecycle-certification.py passed all ten real phases, including source/imported count and query receipts, matching result hashes and retained/peak storage checks.
• Workspace Clippy, formatting, make pre-push-fast and make gate-registry-check: passed on the final tree.

#951 remains open for S20 and conditional S22 host outcomes using #900's ladder. This PR supplies deterministic implementation and admission proof and does not claim host-scale qualification.

Exact-head CI: Test Suite 33954226141 passed at de630749b9b19e99bbdbd38f10bc732727a51105; CI Gate was SUCCESS, merge state CLEAN, and no review threads were unresolved. Squash-merged as 417b6466c027730c0bcbcabb56ac87194585b4fd; #1116 closure verified.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Team

Run ID: bb17b643-d646-4921-a9ab-de491330ef2d

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

Warning

Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}