[edgepillar]

The existing executable corpus targets go-zenon 5cb and does not qualify the later reclaim locks. Add a separate corpus pinned to go-zenon 5a2593ab0d5d56ab15ad7ae7a910323a1565969f, preserving historical corpora and recovery-journal source pins. Check the complete 500-file upstream inventory, retain explicit inverses for all four legacy label descendants, and bind the keyless input and Go guard to the same core. Add independently calculated reclaim-message, ABI and selected memory-VM expectations.

The joint runner prepares one 521-file source view and one fresh compiler workspace for 21 profiles, 556 parent/subtest outcomes and 25 Go commands. Retention selects complete original streams and finite comparisons explicitly, including incomplete attempts. Add a separate current-core CI job and artifact family while preserving the existing jobs.

Validation: all 4,358 local offline cases passed without skips or resource warnings. One fresh Darwin Go 1.23.12 execution passed all 21 profiles and 556 outcomes, with 42 original profile streams. The 53 explicitly selected original files passed value checks and 24 altered-copy refusals. Source and output disclosure checks passed. The selected Darwin SDK contents and transitive compiler provenance remain unauthenticated; these results establish selected finite behavior only.

Fresh hosted PR validation and its original artifacts remain separate gates. Main integration requires that acceptance, and push-triggered main CI requires its own acceptance. Application/core remain NO-GO; complete account admission, independent construction review, signed-account custody, canonical observation, finality and recovery admission remain unfilled. No application signing, node activation, wallet access or broadcast is introduced.

[mrstockholmz]

Verified locally against the published pins:

• Source identity: git rev-parse 5a2593ab0d5d56ab15ad7ae7a910323a1565969f^{tree} in zenon-network/go-zenon gives 43c8eeaee196a9206da7669b0250a000fba9b776, matching the declared tree in NOM_DEV_5A2593.md. The commit is on origin/dev (carries merged Allow reconciliation crash fixtures time to reach exact checkpoints #214, Handle public checker output failure and focus process checks #216, Select focused public-client PR checks with a conservative full fallback #217, Bound Python public IO and account for the pinned dependency README #218), not yet on master — consistent with the doc's "current-source" framing.
• Manifest pins: all 7 SHA-256 pins in nom_dev_5a2593_reclaim_manifest.json recompute correctly from the checked-out files (inputs, values, legacy manifest, Go probe text, qualification module, ABI boundary module, pytest module).
• Message expectations: all 18 reclaim_message_hex values equal Python hashlib.sha3_256 over their declared preimages, so the values file is internally consistent with the Python-side construction (which, per the doc, is never passed to Go). Case counts match the doc: 18 message + 18 send + 40 VM = 76.
• Scope discipline: the values file keeps application: NO-GO, core: NO-GO, independent_assessment: UNFILLED, and the doc states native execution of the combined 21-producer candidate is "not yet qualified". The corpus claims encoding and selected memory-VM behavior only, and says so precisely.

The declared payout-failure VM cases (lowering the synthetic contract balance by one unit and requiring exact storage/balance rollback with a failure receive and no payout) are the most valuable additions — once the combined native run is accepted, they will cover the receive-failure path that a happy-path corpus would miss.

Solid incremental corpus, honest about what it does not establish.

— mr.stockholm.Z