P10.9A: safe independent BRCB triage for cross-slot operator diagnostics

[masarray]

Evidence from 2026-10-11 BCUGE same-PC GE simulator (operator supplied, not committed)

Open SCL and IP Discovery reports were captured from the P10.8E PR merge-test executable f4b07d454215f14dc659e2e9e57e3cfea62b4e98 (parents P10.8D and P10.8E); locked ARIEC 00ad2819b99ffe6b3f885e59aa24a6afb2b4d8e1. Identical static semantic fingerprint 439d4dc0507019533b154ed68f7f4007f39f6be92b5cd3f413ff3b84f619727e; 8 DataSets / 124 exact members / 67/67 displayed points / 6/6 routed RCBs / zero cyclic MMS process polling in both runs. Open SCL: 49 reports, 13 findings / 10 SqNum discontinuities / 3 BufOvfl. Discovery: 55 reports, 7 findings / 4 discontinuities / 3 BufOvfl. RCB slots DIFFER: Open SCL CSWI/EVE GW, Discovery CSWI/EVE SOGI. These cannot be paired as reliability A/B. EVEGGIO has only two GI+overflow frames and 0 post-anomaly follow-up in both sessions.

Change (isolated offline-only milestone)

Preserves the original strict default P10.8E comparator, including rejection of different BRCB slots or DataSets. Adds opt-in --independent-triage using the same existing parser, with separate per-session BRCB records for frames/BufOvfl/findings/SqNum discontinuities/GI-context, later actual reports, bounded retained chronology, and last-anomaly delay from the recorded RCB start-return (receive-time observation only). Marks same_slot_dataset_scope=false and paired_comparison_blocked_reason for cross-slot captures. Null remains unknown if evidence missing. Rejects impossible GI/anomalous-frame counts. All outputs remain NOT_PAIRED_RELIABILITY_COMPARISON; NOT_WIRE_ORDER_PROOF; NOT_SOE_LOSSLESS; NOT_R10_PHYSICAL_ACCEPTANCE.

Usage (Windows, two operator diagnostic files kept private)

python .\scripts\compare-p10-brcb-sessions.py .\OpenSCL-Diagnostic.txt .\Discovery-Diagnostic.txt --independent-triage --output .\P10-BRCB-IndependentTriage.json

Omit --independent-triage for the original strict same-slot comparator (different slots continue to hard fail).

Only 2 existing offline Python scripts changed; original 13 fixtures + 6 new regression fixtures = 19; canonical Windows build already runs them. No new workflow, no engine, runtime/decoder/RCB/MMS/GI/control/GOOSE/polling changes. Baseline P10.8E staging b129e7da32d804effff195497359f5ff0eb56a07, checkpoint unchanged until exact SHA CI passes.

Field causal limitation: no authorized timestamped MMS wire capture supplied; this milestone prepares usable observational evidence, not a fix for the observed historical BRCB overflow or a determination of client/server cause. Follow-up under #520 requires genuinely free identical slots, simulator-only safe non-control changes and permitted packet trace. Preserve P10.6 #518, keep #512 draft, no main/public release.