[masarray]

Based on the installed field build 99d39c7 (P0-E + viewport stack), preserving its report causality and command flight authority.

Opening the Command Dock previously forced inspection of every selected control, including verified status-only targets. Inspection invalidates descriptors and queues MMS work. Reopening now uses existing verified state, while explicit Refresh still requests fresh inspection. Busy targets are excluded; inspection ownership is claimed before local throttle wait and cleaned up on cancellation.

Pins ARIEC61850 PR #183 at c85d4e20db722bd0c2729f2533b323b13d3ef735: non-RCB control InformationReports cannot become BRCB frames via single-monitor fallback. Ordered command evidence now exposes service and post-response wait durations for locating the observed 8–13 second native action latency.

Validation: engine compiled with zero warnings/errors; 64 relevant Fact tests invoked directly passed. Standard local test host fails before execution in the container; Windows CI and physical acceptance remain pending. Added inspection policy regression. No claim that the complete physical command delay is fixed. No timeout/interlock changes and no main merge or release. See docs/CONTROL_LATENCY_TRIAL_2026-10-20.md.

[masarray]

Companion command scheduling PR #638 is now stacked on this PR #635 at exact base 23872578ca3d410520039e694567f6c7ad351674. It removes parallel UI admission of read-only dock inspections ahead of native serialized MMS control, yields to WPF input between inspections, and skips new inspection while a command is active. It does NOT replace engine #183 command stage instrumentation, or claim the physical 7.96/8.10/13.32-second native latency is fixed. Separate Browser leaf/progress PR #636 has disjoint files. Preserve cross-branch ancestry; test exact integrated head before promoting.

[masarray]

2026-10-20 control audit: follow-up draft PR #640 (stacked #638 -> #635) fixes source-proven Command Dock flicker: Iec61850MonitorDevice.RefreshCommandSignals() ALWAYS did CommandSignals.ReplaceAll, emitting whole-grid Reset even for identical rows; 2.5 s UX preload timer invokes refresh on zero candidates, and DistinctOperable changes with transient proof. PR #640 implements in-place row reconciliation and regression; full design is docs/COMMAND_DOCK_RELIABILITY_ARCHITECTURE_2026-10-20.md. Also repaired original P109S fixture in this PR #635 HEAD 2aa0461d... after exact CI log showed an invalid CO identity—not a product guard defect. Important: full native 8–13 s CommandTermination diagnosis remains open, must use engine #183 stage traces and real bench before performance claims. Leaf/progress sibling PR #636 touched Models/MonitorModels.cs in a different section and requires semantic combined-head integration, not a stale overwrite.

[masarray]

D28 archival on maintainer direction: older experimental IEC 61850 branch superseded by the latest successful, operator-accepted baseline. The operator will not recapture evidence for this historical trial. Closing as NOT MERGED, without modifying runtime or stable release. Historical HEAD 2aa0461d672d4e554ba4f61c831dcd1bb03c4d63 and its distinct code/tests remain preserved in GitHub; those unmerged changes are not falsely claimed as implemented. This retirement removes a stale integration proposal, not an established safety limitation or protocol fact. No follow-up capture required for PR cleanup.