Trial: repin ARSAS 1.6.36 to green ARIEC convergence head

[masarray]

Purpose

Qualification-only repin of the existing ARSAS 1.6.36 trusted-SCL golden-wire lane to the exact ARIEC convergence head that passed final ARIEC code gates, plus correction of the field-trial contract so operators test the actual one-shot GI behavior.

Exact ancestry

• Base ARSAS field-trial head: 131a17d87e40e379050b30c1497c50c5c4af647e (PR Trial: trusted-SCL golden-wire runtime on ARSAS 1.6.36 #310)
• Previous ARIEC trial pin: d50e5bcb9fd428fe3d80ac72f8d4015a575cfda5
• New ARIEC convergence pin: 0023ef9a4373855497464ed3979e359c4041c95d
• ARIEC source PR: Diagnose v1.6.20 release publication #132
• ARIEC exact-head .NET CI #603: PASS
• Current ARSAS integration head: 136d547da73b1c6f062820aa4021420bc7b26c8d

Scope

Two non-runtime files change relative to PR #310:

1. engines/ARIEC61850.lock.json
2. docs/SCL_GOLDEN_WIRE_PHYSICAL_TRIAL.md

No ARSAS application code, UI, FAT, COMTRADE, Smart Control, packaging logic, or report workflow code is changed in this PR.

The lock deliberately keeps ref: main because both Build ARSAS and Validate ARSAS Windows installer enforce that metadata invariant before checking out the immutable commit SHA directly.

The structured fieldProvenBaseline block is preserved verbatim from the canonical PR #310 lock. ARSAS regression tests intentionally treat the documented PR #76/#77/#78/#79/#80/#81/#84/#85/#86/#87/#88/#89, G1/G2/P0/P1 and physical BRCB ancestry as part of the compatibility contract, not disposable prose.

Preserved runtime contract

The new pin preserves the trusted-SCL ARSAS path already implemented in PR #310:

• verified SCL remains authoritative for DataSet/RCB identity;
• no hidden live DataSet-directory browse or dynamic DataSet mutation;
• receiver registered before report-control writes;
• URCB: Resv -> RptEna when Resv is exposed;
• BRCB: direct RptEna, with ResvTms only as retry compatibility after actual direct-enable rejection;
• two whole-RCB verification reads;
• one-shot startup GI=true only after routing is registered and activation succeeds;
• GI rejection fails closed and cleans up RptEna/reservation;
• steady state remains event-driven with no cyclic MMS process polling.

The convergence also carries the explicit buffered-BRCB latest-state regression guard and preserves SCL RptEnabled@max as diagnostics metadata only; it is not used to synthesize runtime RCB names.

Regression discovered and fixed during qualification

The first full integration regression compiled successfully but exposed 7/1133 test failures. All seven were provenance-contract assertions against historical lock text that had been shortened in the temporary pin (PR #87, PR #89, dynamic-attempt, P1 hardening, exact G1 SHA, etc.); no engine/API/runtime test failed.

Head 55b7655... restored the full canonical field-proven lineage while retaining the new immutable convergence pin.

A second audit then found the physical-trial document itself was stale: it still referenced an older engine and stated that GI was not sent. Head 136d547... corrects the documented physical contract to the actual PR #310/convergence behavior: receiver-before-write, RCB activation/readback, one explicit startup GI, fail-closed GI cleanup, event-driven steady state and deterministic release.

Exact-head software gates

Only workflows on 136d547da73b1c6f062820aa4021420bc7b26c8d count for final qualification:

• Build ARSAS #2822 — queued;
• Validate ARSAS Windows installer #975 — queued;
• Validate IO List Testing #1365 — queued;
• Validate SV evidence bundles #1804 — queued.

Build ARSAS includes full application regression tests, real portable single-EXE publish and startup smoke. Installer validation includes restore/build/tests, native ArdIrec bridge integration, installer compile, silent install/runtime smoke, uninstall, checksums and artifact creation.

Required gates

Keep Draft. Do not merge this PR or PR #310 automatically.

1. All four exact-head software workflows above must PASS.
2. Promote the exact green pin into the canonical ARSAS field-trial lane only after software results are reviewed.
3. Physical IED regression must confirm BRCB + URCB startup, all initial values, structured values, event-driven updates with no polling, and deterministic RCB release.
4. No production merge/release without explicit approval after those gates.

[masarray]

Superseded by Draft PR #314. PR #314 starts from exact green PR #312 head cd1e172ff59160d7b7b1875008b515817fe180ab and applies only the convergence engine lock + corrected physical-trial contract, so it preserves the active real-IED-feedback/analyzer workstream instead of qualifying a stale #310-only base. This PR remains useful as isolated protocol-repin regression evidence, including the lock-ref and provenance-lineage failures found during qualification, but it must not be merged.