Batch 361-A (TRUTH DETERMINATION, DOC-ONLY: the prescription's windows are EXISTENTIAL WITHOUT TAIL-VANISHING, so the field is not determined -- MM.FLM861Book.isVertexOperator_unique is MARKED SUSPECT and NOT attacked).

Before spending a batch on the piece registered in batch 360-A, its truth was checked -- and the check FAILED it.  REASONING ONLY (no Lean witness): this is a SUSPICION of record, NOT a refutation, and it is marked and left un-attacked per AGENTS.md ("When a registered sorry is refuted", section 1).

Finding.  IsNormalOrder and IsBaseField state their sums with an EXISTENTIAL window and do NOT require the terms beyond it to vanish.  The book's content is exactly that finiteness ("the sums are finite on every vector"), whose value is the LIMIT: a window-relation without the tail conjuncts also admits SMALL windows, and the increments at N are X (-N-1) (Y (m+N) w) (first sum of (8.4.16)) and Y (m-1-N) (splitPlus X N w) (second) -- the creation part of the derived factor against a fixed value, generically non-zero below the truncation thresholds.  So the relation pins the value only to the set of partial sums, and the determinism route sketched in batch 360-A's docstrings cannot be carried out as spelled.  (Whether a concrete datum really exhibits two different partial sums is NOT compiled here; that probe is the natural first step of the repair batch.)

Landed (no statement changed anywhere):
* Definitions/FLM861BookDefs.lean: a WINDOW DEFECT OF RECORD note on IsNormalOrder and on IsBaseField (the window is existential, the tails are unconstrained; the missing conjuncts are the book's finiteness, and in the honest case they are provable: the p-tail is the PROVED annihFam_killsAbove, the q-tail is the lattice confinement of azPowD);
* Definitions/MMFLM861BookUniqueness.lean: the SUSPECT mark on isVertexOperator_unique's docstring, naming the repair and forbidding the attack until it lands;
* NOTES.md section 278 (the finding, the ledger consequence, the proposed repair and why the repair is not a weakening: it strengthens the HYPOTHESIS IsVertexOperator of the three frozen statements, exactly as the book's finiteness does, and the honest field satisfies it -- that IS the producer's tail bookkeeping);
* HANDOVER.md section 3 (the suspect entry) and section 6 (the batch record and the repair's shape).

Repair (the next batch's shape, a definitional repair of the 344-A/345-A family, NOT a weakening): make the window a genuine TRUNCATION POINT -- add the two tail conjuncts to IsNormalOrder's conclusion and the p-/q-tail conjuncts to IsBaseField's rectangle; after that the batch-360-A route works verbatim.

Evidence: lake build FinalCheck EXIT=0 (9175 jobs) with the same NINE 'declaration uses sorry' warnings; the three book-facing statements byte-identical (3 blocks, 0 mismatches); forbidden_tokens on the touched files: admit=0 axiom=0 native_decide=0, sorry only where registered (the two 8.6.25/8.6.27 halves, this module's two pieces, #2, moonshine, FLM889Book, FLM8102Book x2).
