[igorls]

A fresh admission can change the same pending projection's diagnostic reason after an earlier detached logical recovery view was read. The old worker qualifier could then reject the physical snapshot despite preserved source/binding state. A controlled replay reproduces that mechanism; the historical service interleaving remains unknown.

Sample fresh recovery, unavailable API behavior and the complete physical audit within the original returned isolation deadline. Retain every rejected bounded sample and its original observations/error causes. Retry only a complete sample-bound admission/reason transition; other invariant, timeout and capture failures stay fatal. Qualification recovery reads and supported API/Mongo requests use the remaining budget. Backend admission/CAS, source authority and serving behavior are unchanged.

Validation:

• Native Node 24.19.0 and 26.4.0 each pass all 51 focused controls in independent review and the replayed integration checkout. Application/qualification checks, builds and existing physical-audit tests pass.
• All nine final-head CI jobs pass. Both Linux build legs report 1,586 passed / 1 optional skip / zero failures / 10,609 assertions, and execute the 51 new native controls.
• Focused real MongoDB 8.0.34 checks pass serially on Hyperion-3 with both native runtimes: raw child exit 0, complete admission-transition then passing audit samples, and verified owned-resource cleanup. Root rehashed all six archive chunks and 102 payloads, checked both primary receipts/reports and all protected guards. Emission hash receipts cover 535 files; emitted bodies were not retrieved.
• Independent complete-delta Spec: zero findings. Standards: zero rule violations and one nonblocking canonical-comparator duplication suggestion. Literal patch replay matches reviewed tree f9d5b5b4.

The initial private wrapper failed after the helper ran but before retaining its child exit. That attempt remains unaccepted; its evidence and both wrapper versions are preserved. Fresh uniquely named cases supplied the accepted results.

The real-server control uses a constructed empty-mutation prefix with retained source IDs. Historical cause, durability qualification, fresh full Linux 1..1480 worker replay and Wire/release acceptance remain separate under #191/#277.

Refs #341, #191, #277 and eosrio/hyperion-tools#24.

[igorls]

Independent review of the complete ten-file delta at 40d589a0 (tree f9d5b5b4, identical to reviewed 20f36111) against c885ee41.

Standards

Zero documented-standard violations. One nonblocking P3 judgement: the sampler repeats the auditor's canonical comparator; a small shared qualification helper could prevent divergence. This remains a recorded cleanup suggestion.

Spec

Zero findings across all ten files. The original isolation deadline covers fresh logical/API/physical samples; only a complete sample-bound admission transition retries. Independent invariant, timeout and capture failures remain fatal. Earlier rejected samples are retained.

Both reviewers independently passed all 51 controls on native Node 24.19.0 and 26.4.0, including the preserved intentional failure. Root hash-verified the reports/evidence, literal patch replay, fixture identities and its own 51-case runs, typechecks/build and physical-audit tests.

All nine exact-head CI jobs passed. Both Linux build legs report 1,586 passed / 1 optional skip / zero failures / 10,609 assertions across 1,587 tests and 143 files; both also execute the 51 new native controls. The focused Hyperion-3 actual-auditor check remains pending: an initial post-run wrapper failed before retaining the original child exit, so that attempt is not accepted despite its passing helper report and verified cleanup. Fresh cases preserve the failed wrapper evidence.

Source/component review does not establish historical causality, full Wire replay or release acceptance.

Summary: Standards 0 rule violations, 1 nonblocking P3; Spec 0 findings.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.

An organization owner can choose the repository organization as the billing source for reviews requested by members with a paid Copilot license. Learn about Copilot code review billing. Review the Copilot code review billing settings.

[igorls]

Post-merge verification: f4df8c44f11016206326950cf1157d67a584d958 has the exact reviewed/tested 40d589a0 tree, and all nine separate parent CI jobs now pass. #341's completed outcome records the fresh passing actual MongoDB controls on both native runtimes, verified cleanup and independently verified archive/all 102 payloads. This completes the scoped sampler; parent #289 remains draft and full Wire/capacity/recovery qualification remains open.