feedback5 R19D: 5-阿戈尔 revives — the members the battle-start devour knocks out revive once every devour pass is over, by board position (columns from the left, bottom first: the [ASSUMED] reading of the community report's 「左下到右上」), not in knock-out order; later knock-outs take the slots left in time order, uneaten members none (community report relayed 2026-10-07); golden: unchanged

The report's steps 1-2 were verified and hold: the bond HP is on every member before the devour (an elite 深巡,
2904 HP / DEF 655, stands its 4345 from 15 layers on), and each mark is one 5000 物理流失 less DEF (PRTS), so a member
in the middle of a chain is marked by every 阿戈尔 behind it — that is what knocked out the reporter's 蓝蒂 / 小鲨.
Golden unchanged: the corpus's only 5-阿戈尔 battle (bond-egirShip-high, 999 layers) has no devour knock-out.
DESIGN §25.22.5.

docs/history/0.1.4.md

@@ -74,7 +74,7 @@ Reports after the 0.1.3 release. Each was checked against the official data and
   高台; the bots plan each record on a 高台 that covers the road, a 重装 on the road, no 高台 fallback for it);
   `test/ui/gameLogic.test.js` (the drag mirror under four loadouts).
 
-### 24.3 阿戈尔 5: the 3 revives go to the first 3 members knocked out (GitHub #105, #140; the owner's decision of 2026-10-05) — `content/bonds/core.js installEgir`
+### 24.3 阿戈尔 5: the 3 revives go to the first 3 members knocked out (GitHub #105, #140; the owner's decision of 2026-10-05) — refined in 0.2.0: the battle-start devour's knock-outs revive by board position once the devour is over, see §25.22.5 — `content/bonds/core.js installEgir`
 
 - **Seen**: GitHub #105 (comment 2) and #140 (comment 3), and a report the owner quoted: 「没被吃的阿戈尔干员也会占用复活名额」
   — with 5 阿戈尔, "whatever the food chain eats, only 3 阿戈尔 stand at the end". #140 item 3 (screenshot): the chain ends

docs/history/0.2.0.md

@@ -1980,3 +1980,66 @@ here, each reproduced first; the rest were answered (not a bug, already fixed in
   the parent), `test/match/playtest6-matchflow.test.js` #235 (2; the 联防 one fails on the parent: losses = the LP
   settled, no result ticker, gone in the next phase), SP_E2E `test/ui/resultbox.e2e.test.js` (3, mock variants
   `unite[,through][,helper|,dead]`). Docs: PLAYING.md §1 / §5, design/network.md, design/client.md, design/match.md.
+
+#### 25.22.5 「阿戈尔盟约的二阶复活逻辑应该有些bug」 (a community report the owner relayed on 2026-10-07, two screenshots not yet seen): the 5-阿戈尔 revives of the battle-start devour's knock-outs go by board position once the devour is over — `content/bonds/core.js installEgir` (`egirReviveScan`, `egirStart`)
+
+- **Seen**: the player's model of the official game: 「原作里是第一步给所有鱼加盟约的血量，第二步是开始一个个大鱼吃小鱼，
+  第三步是检测左下到右上哪些鱼睡着了给他复活」 — a fish that survives the devour on the bond's HP takes no revive; in the
+  remake 「蓝蒂的位置肯定是比水月要更优先复活的」 and 「中间位的小鱼（蓝蒂、小鲨）已经超5000血量了还是被吞死了」; in an
+  official game they recorded an elite 深巡 (under 3000 HP in prep) stood its devour, 「所以肯定是部署后先经过盟约加血才
+  活了下来」. Reproduced on 0255afd (real chess, scratchpad w5-r19d-notes/repro.mjs): (a) one row 浊心斯卡蒂 → 乌尔比安 →
+  斯卡蒂 → 幽灵鲨 → 水月 → 深巡, all elite, 60 layers: 斯卡蒂 at 7513 HP and 幽灵鲨 at 4965 each take two marks (浊心斯卡蒂's
+  and 乌尔比安's, 2 × 4751 and 2 × 4635); 斯卡蒂 is saved only by her module DRE-Y. (b) 乌尔比安 → elite 深巡 (2904 HP, DEF
+  655), one mark of 4345: knocked out at 0 and 14 layers (3920 / 4327 HP), standing at 15 (4356 → 11 HP). (c) Two rows,
+  乌尔比安 → 水月 → 海霓 on the top row and 浊心斯卡蒂 → 幽灵鲨 → 深巡 on the bottom one: four knocked out at t = 0 in the
+  order 水月, 海霓, 幽灵鲨, 深巡 — the first three revived, 深巡 down.
+- **Official**: PRTS 卫戍协议：盟约 下半/PRTS盟约记录, 阿戈尔 (fetched 2026-10-07, unchanged since the 2026-10-03 copy): the
+  3-tier gives the members 生命值 +(35+1×层数)% (bonds.json `base_max_hp` 0.35, `max_hp_per_stack` 0.01; 直接乘算, the
+  page's 「盟约效果…提供的属性加成均为直接乘算」) and the devour; the 5-tier the revives (`max_free_respawn_cnt` 3). The devour
+  is 「5000点物理流失」 (the in-game text says 物理伤害; PRTS corrects it), one per mark (「每个标记令目标受到一次5000点物理流失」),
+  and the marks chain (「以及被自身吞噬的【阿戈尔】干员身前一格干员」), so a member is marked by every 阿戈尔 behind it in its
+  chain. PRTS 作战机制 §伤判效果与生命流失: a 物理流失 「会受到目标当前防御力…影响而相应衰减」 and skips every other step
+  (no 脆弱 / 庇护, no dodge, block, shield or barrier). No source gives the moment of the HP bonus or the order of the
+  revives; PRTS 卫戍协议/帮助: 「满足再部署条件时…自动部署」.
+- **Steps 1 and 2 hold (not changed)**: the bond's HP is a match-long buff put on the members when the battle is built, so
+  each is deployed with it at full HP and the devour finds it — (b) is the player's official 深巡 exactly: it stands from 15
+  layers on. The prep card's 「开战时」 生命上限 comes from a started preview battle (match/intents.js `_unitStatsOf`), so it
+  includes the bond's HP as the battle does. The devours resolve one by one in marking order, one 5000 物理流失 less DEF
+  per mark, as PRTS says; a fish over 5000 HP stands one mark and falls to a second — that is (a)'s 「中间位的小鱼」: in a
+  chain each 阿戈尔 behind it marks it. So the HP was real; what killed 蓝蒂 and 小鲨 was the second mark.
+- **Cause (step 3)**: §24.3's rule gave the 3 slots to the first 3 members knocked out, at the knock-out, so inside the
+  devour the knock-out order decided — the food the first marks reach (often the far end of a chain, or the row whose big
+  fish marks first) took the slots while a member further left fell to a later mark and stayed down.
+- **Now**: the members the battle-start devour knocks out wait until every player's devour pass on the field is over (the
+  window `egirStart`: the passes counted at install, a depth while one resolves, the 5-tier players' batches run after the
+  last pass); then those still down are revived by board position (`egirReviveScan`) while slots last; the rest stay down,
+  their first knock-out spent. Unchanged from §24.3: a member takes one slot at most, on its first knock-out; a member never
+  knocked out takes none (survivors = uneaten + up to 3 eaten); a knock-out later in the fight takes a slot left at once,
+  in time order; a save by its own kit / item / band is no knock-out; a member entering 联防 down is not knocked out; each
+  battle counts its own; a mirrored boss player scans its own board. (c) now: 幽灵鲨, 水月 and 深巡 revive (left column
+  first, bottom first), 海霓 stays down. A member a teammate's devour knocks out on a shared field waits for its owner's
+  scan too (before: back at once, so it was standing — and marking — in its owner's pass).
+- **Changed from §24.3**: only which battle-start knock-outs take the slots and when they come back (after the last devour
+  pass instead of inside it). With three or fewer knocked out at t = 0 the same members revive; the order of their
+  redeploys (the later deployed is attacked first) is now the scan order.
+- [ASSUMED]: (1) 「左下到右上」 on the player's own board as seen in prep — row 0 at the bottom of the screen
+  (render/projection.js), col 0 at the left, a mirrored player from its own left as `egirOrder` — read as the columns from
+  the left, bottom first inside a column, like the act-2 positional orders (the deployment, the devour order). The other
+  reading, the rows from the bottom with left first inside a row (act 1's deployment order), is the one-line comparator in
+  the comment above `egirReviveScan`; `test/content/bonds_core.test.js` '阿戈尔 5 (community report, 2026-10-07) step 3'
+  pins the choice — under the other reading its F1 stays down instead of F4. The screenshots should settle it. (2) Inside
+  the window a devour knock-out is otherwise any knock-out: 被击倒 triggers fire and 不屈 rolls (before, the 阿戈尔 revive
+  answered first and 不屈 never rolled); a member 不屈 redeploys is awake at the scan and takes no slot (「哪些鱼睡着了」).
+  (3) Each player's devour stays its own pass (§24.4 [ASSUMED] (2)).
+- Golden: unchanged — traced with an instrumented copy: the corpus's only 5-阿戈尔 battle (bond-egirShip-high, 999 layers)
+  has no devour knock-out, and no 5-阿戈尔 member is knocked out later in any scenario.
+- Tests: `test/content/bonds_core.test.js` +3 — step 1 / 2 (the elite 深巡 at 14 / 15 layers, the HP before the devour; an
+  elite 幽灵鲨 at 5013.54 HP stands one mark and falls to a second in a chain; passes on the parent too: those steps held),
+  step 3 (four food knocked out in the order F1, F2, F4, F3: all marks first, then F3, F2, F1 revive and F4 stays down —
+  normal, 联防 with a marker entering down, the mirrored boss player; the parent revives F1, F2, F4 inside the pass), later
+  knock-outs / uneaten members / 不屈 (fail on the parent); three titles of the older 阿戈尔 tests reworded.
+  `test/content/feedback4-egir-unite.test.js` +1 (a teammate's mark knocks out p2's 幽灵鲨: revived after p2's own pass,
+  fails on the parent). Docs: research 02 §3.6 and its JSON spec (data/bonds.json `egirShip.spec` regenerated with
+  `build-data --offline`; only `tiers[0]` / `tiers[2]` / `algorithm[3]` / `assumed[0]` changed — the stale assumption
+  「5000 physical 流失 ignores DEF and shields」, wrong since 0.1.1, gave way to the scan reading), DESIGN §5's `death` row
+  (design/engine.md), docs/SIM.md (`death` hook), the core.js comments; §24.3's title points here.