[masarray]

Field regression 2026-10-20 — P0-E (DRAFT, never release/merge without evidence)

Operator provided Windows diagnostic app 1.6.40+a37ab2c9 (sibling of P0-D #630 HEAD) and screenshots: discovery took 16.526s (directory 10.166s, types 3.054s, projection 2.856s), some analogue / SPS points remained Unknown/RCB pending despite exact single ingress Buffer01, and dchg events were described 'Spontaneous' despite their buffered RCB channel.

Focused non-naive changes

• Index authoritative canonical signal references once per projection flight, preserving the existing case-insensitive NormalizeReference equivalence and first-observed precedence; remove quadratic signals.Any fallback duplicate checks for LN targets and mapped MMS fallback. No new process value store or MMS I/O.
• Map IEC dchg ReasonForInclusion to Data change, not an unsupported transport/replay assertion. Keep exact RCB acquisition, RptID, raw reason, and verified buffered replay as independent dimensions, with existing tests updated.

Qualification

CODE pending exact-head Windows CI. Performance improvement is algorithmic, NOT a measured field speedup until same-endpoint repeat-run performance capture. Unknown/RCB-pending root cause, Report stream alias, GI chronology and 8.2s control feedback remain open field/engine investigations. No control or RCB writes added, no implicit MMS polling, no main merge, no release. Based on PR #630 which stacks on #628 and engine #181 exact pin.

[masarray]

D27 backlog reconciliation — historical PR branch fully contained in integrated RC. Exact GitHub ancestry proof: 0a7e555df10af884f4f777c21f29b864d734f26c..d4598619b1b8fc21487cf5beb83a8a8284d0e9cf has ahead_by=863, behind_by=0. Thus no commits from this branch would be lost by closing its redundant PR. This is closure as absorbed ancestry, not a direct PR merge. RC #727 holds the code; blocked main promotion #729 and live field/engine gates remain open. Do not treat this historical closure as IEC 61850 physical acceptance.