[masarray]

Stacked lifecycle milestone for #446, based on P7.1 PR #459.

Why

The 7SX85 physical run proved the reporting data plane, but it also exposed a lifecycle weakness: the WPF shell still used ICMP Ping as an authoritative death signal and owned a second reconnect state machine on top of the per-IED MMS runtime.

That can falsely tear down a healthy IEC 61850 association when ICMP is filtered/rate-limited/transiently lost. It also made post-reconnect report proof vulnerable to stale HasReportStream state from the previous association and used a fixed 3 s delay for proof presentation.

Changes

• remove UI-owned ICMP death detection and duplicate reconnect ownership from MainWindow.P0AssociationLiveness.cs;
• keep MMS association/reconnect ownership exclusively in Iec61850MonitorRuntime, which already has bounded client cleanup, reconnect budgets and report re-arm staging;
• clear HasReportStream / report pulse whenever an association is marked offline and again when a replacement association installs;
• make InformationReport proof explicitly association-scoped;
• replace the fixed Task.Delay(proofWindow) presentation gate with an event-driven wait on HasReportStream property change;
• retain the bounded proof window only as a timeout/diagnostic boundary, not as a polling mechanism or activation prerequisite;
• log observed first-report latency when traffic arrives;
• correct static-report-only reconnect diagnostics so they never claim MMS polling resumed when process polling is intentionally disabled.

Safety

• no change to ARIEC RCB/DataSet activation semantics;
• no vendor branching;
• no process-value heartbeat is introduced;
• no polling fallback is introduced for Static DataSet report-only mode;
• absence of report traffic within the diagnostic window remains armed/report-pending, not an activation failure;
• reconnect cleanup remains bounded by the existing SmartReconnectPolicy.ClientCleanupBudget.

Regression

Adds P7.2 guards proving:

• no Ping/ICMP authoritative liveness or second UI reconnect owner remains;
• report traffic proof is cleared on offline + replacement-association boundaries;
• static traffic proof is event-driven and has no fixed Task.Delay(proofWindow);
• Static DataSet reconnect diagnostics preserve the no-polling invariant.

Physical retest is still required after software CI.

[masarray]

D16 historical reconciliation — close as integrated in protected RC ancestry, not merge into main.

• Historical exact HEAD ccf692101844aeeb780e6758882f78440bfc00a4 → RC 817c3910e8223fa2ac531a4426ada5d2d4bbbe50: GitHub compare shows behind=0 (ancestor), all 5 original changed paths still exist in the current RC. No old unique branch commit is required to import.
• Byte-for-byte matching current RC Git blobs: MainWindow.P0AssociationLiveness.cs; tests/ARSAS.Tests/P0LatestFieldRegressionTests.cs; tests/ARSAS.Tests/P72AssociationLifecycleRegressionTests.cs.
• Shared source has since evolved and is not asserted byte-identical: MainWindow.IoTesting.AutoConnect.cs, Services/Iec61850MonitorRuntime.cs. Source checked for original assertions/behavior: the WPF AssociationLiveness file is byte-identical and has NO ICMP Ping second reconnect owner; current RC retains event-driven WaitForReportStreamAsync, association-scoped InformationReport proof clearing, and safe reconnect with Static DataSet process polling disabled.
• Original exact regression test(s) remain in RC. RC exact-head Windows Build, R7 SCL and interoperability reference guard are SUCCESS; physical FAT is a separate acceptance gate.

Decision: old stacked PR is redundant as a merge unit after source/test/behavior mapping. GitHub audit history is preserved; retain overall #446 unresolved physical and release gates and protected #727/#729. No main merge, engine repin, stable release mutation, or bypass of the acceptance guard.