Update 2.27

v2.27.0BETASeymour

AURORA: structural GROW of chunk3 (byte-exact ADD/REMOVE of a monster) — the "chunk3" blocker is gone

• Answers HANDOFF_AURORA_CHUNK3_RESOLVE_2026-06-07. Before, only MOVING an existing anchor was possible (v2.12.0); now the monster COUNT can change. Packing-model RE (corpus 863, proven): the area-record packs its 8 arrays in a FIXED pointer order (origin<party<partyB<aeon<monA<monLive<monB<camera, 863/863); capacity = next-pointer − pointer; monLive(+0x20) is TIGHT (==MonsterPositionCount@+0x06, 856/863) while monA(+0x1C) reserves ~12 and monB(+0x24) reserves 3..78 → the reserves absorb a higher count without their own resize. The "mystery gaps" were the reserve; +0x30..+0x5C = 6 area/camera floats (not pointers). BattleArenaGrowWriter.GrowMonsters = an insert/remove splice at monLive's end + re-stamp of only monB(+0x24)/camera(+0x2C) + the chunk table + +0x06; guards single-area + tight + reserve ≥ cap (HardActorCap=8 conservative; data max=15 znkd09). Gate BattleArenaGrowLab PASS (in offline_ci.ps1): RT0 no-edit 700/700 byte-identical, ADD +1 696/696, REMOVE −1 393/393 (clean re-decode). Ineligible battles refused honestly (149 multi-area + 6 non-tight + 13 anomalous-reserve → still editable via MOVE). Battle→scene transform reconciled = IDENTITY (handoff 2.3). Doc docs/reverse/FFX_AURORA_CHUNK3_PAYLOAD_MODEL_2026-06-07.md. Build 0 errors. Honest: offline byte-safe (RT0) ≠ the game accepting the new count — in-game UNPROVEN (RT2/probe + owner OK; exact spawn-loop ceiling = IDA pending)