P0: recover Build #1888 FAT live and shared SCL authority

[masarray]

Baseline

This P0 recovery branch starts exactly from the accepted golden ARSAS Build #1888 source:

• ARSAS: 712e2c557d5da83d2f02b81f3a58905e985d1e50
• Golden EXE SHA-256: 063751919b51ef1bf13f666441ec2fff7e1f5a54a5b647295c158abbd732a0d6
• ARIEC61850 lock remains: 11ab2304482600c19ba979f4fc9021ddb46b9af9

This PR does not use Build #1868, PR #257, or Build #1898 as its recovery base.

P0 fixes

• Project the existing _runtime.PointUpdated stream into FAT Runtime.CurrentValue without UI polling/timers or per-cell subscriptions.
• Canonicalize Boolean FAT presentation to True / False before first visible FAT frame and on subsequent runtime updates.
• Keep analog evidence owned by the existing FatAutoCaptureCoordinator; hide the normal ✓ Capture row button while retaining Recapture correction behavior.
• Initialize Engineering control Interlock and Synchronism checks ON once when command signals enter the runtime model; later operator changes are not forced back.
• Preserve an explicit Static DataSet choice when SCL is opened from FAT instead of silently demoting it to Hybrid.
• Preserve Static DataSet authority for appended FAT SCL IEDs as well.
• Mark an already-populated SCL Engineering model reusable so Engineering Play takes the saved/SCL connect path instead of launching a second Live Discovery.

Safety / non-goals

• No VirtualizingPanel.SetVirtualizationMode(...) changes.
• No _uiFlushTimer/DispatcherTimer live mirroring.
• No ARIEC61850 engine change.
• No RCB/DataSet backend redesign.
• No evidence/session-format redesign.
• No FAT legacy-first-paint -> Value1/Value2 schema redesign in P0; that remains a separate P1 technical-debt item after recovery is stable.

Bench acceptance gate

• FAT rows and Command Panel stay populated.
• Engineering/Diagnostics remain responsive through FAT -> Engineering -> FAT.
• No WPF virtualization exception.
• Physical Open/Closed changes are reflected in FAT Live Value consistently with Engineering.
• Boolean values appear as True / False from first visible value and remain stable while scrolling.
• Analog Value1/Value2 captures automatically; normal manual Capture button is absent.
• SCL opened from FAT with Static DataSet populates Engineering immediately; Play connects/activates the existing SCL/static report model without running full Live Discovery.
• Static DataSet/RCB and Command Signals remain present and functional.

Do not merge until CI is green and the exact artifact passes physical relay bench validation.

[masarray]

Bench regression follow-up after Build #1912 is now implemented on head c5fc74e2211c72f23ddb368f9bfb2a76ae2f0df5.

Targeted fixes:

• Runtime PointUpdated FAT projection is now WPF-free/fail-isolated on the worker callback, preventing a presentation exception from turning a physically successful control into COMMAND FAILED.
• FAT live projection resolves exact live-point identity and bridges unique static DataSet DisplayReference members to the ARIEC-resolved scalar runtime ObjectReference.
• IoTestLiveBindingService now performs the same exact structured-member -> runtime-leaf bridge before FAT session arming, so analog rows can become LivePointReady and use the existing 3-sample auto-capture path.
• FAT LIVE/VALUE1/VALUE2 templates render canonical True/False; normal per-cell ✓ Capture controls are removed from the replacement template while Recapture remains available.
• Header actions are split into primary and secondary adaptive rows.
• Close lifecycle quiesces FAT live projection before evidence sealing/persistence and resumes it if close is aborted.

Validation on this exact head:

• Validate IO List Testing #792: SUCCESS
• Validate SV evidence bundles #1017: SUCCESS
• Validate ARSAS Windows installer #700: SUCCESS
• Build ARSAS #1919: SUCCESS
• Build solution: SUCCESS
• ARSAS regression tests: SUCCESS
• Portable publish: SUCCESS
• Portable smoke-test: SUCCESS

Artifact ARSAS-win-x64-portable-single-exe id 9979633455 was produced by Build #1919. Extracted EXE SHA-256: 140f30e4317500cd45c93b614cbb3f9be96f76b1c1c7e42c8d6e4c1e02a786ae.

PR remains DRAFT. Do not merge until the exact Build #1919 executable passes the physical relay bench acceptance gate.

[masarray]

P0 bench follow-up applied on head ed7152f9bfcbead63ae1606c286e6e58420c6273.

Changes in this follow-up:

• FAT Start/Continue now keeps evidence baseline hashing + durable journal I/O off the WPF Dispatcher.
• FAT Stop now runs final journal append/dispose/hash verification plus durable workspace save off the WPF Dispatcher.
• Workspace close uses the same non-blocking StopAll/seal/save path after quiescing the FAT runtime projection.
• Remove from FAT recovery is deterministic again: Removed Signals (N) no longer depends on an exact Button.Content == "Engineering" caption. It also falls back to the stable Export action panel, so a renamed Engineering Workspace caption cannot hide restore again.
• Existing RemovedFatSignalsWindow / RestoreToFat() remains the authority; no signal identity/evidence is deleted by Remove.
• Added P0 regression coverage for responsive Start/Stop/Close and reversible Remove.

CI for this exact head is green:

• Build ARSAS #1924: SUCCESS
• Validate IO List Testing #797: SUCCESS
• Validate SV evidence bundles #1022: SUCCESS
• Validate Windows installer #703: SUCCESS
• portable EXE publish + smoke-test: SUCCESS

Keep PR Draft. Physical relay bench still required before merge.

[masarray]

P0 bench candidate updated at 50d7f899224d61010e87c95f50826594a0fde4bc.

• Restored exact golden ARIEC61850 pin 11ab2304482600c19ba979f4fc9021ddb46b9af9.
• FAT Session.Start/Stop/StopAll remain on the owning WPF Dispatcher; no whole-session Task.Run, preventing the cross-thread different thread owns it Start FAT failure.
• Multi-RCB Legacy SAS export is implemented in the ARSAS layer using the golden engine's existing SclReportControlFilter multi-selection capability; no acquisition/control engine change.
• Independent checkbox selections are preserved; availability/default focus does not clear existing operator selections.
• In use remains selectable with confirmation; no-DataSet RCBs remain visible/selectable as evidence but P0 CID export refuses to invent a DataSet binding.

CI: Build ARSAS #1950 SUCCESS; regression tests SUCCESS; portable publish + smoke test SUCCESS; IO List, SV evidence, Legacy SAS RCB guard, and Windows installer validations all SUCCESS.

Keep PR Draft. Physical relay bench validation is still required before merge.