KB: the authored turn-rate landing recorded -- open-questions Sunday item marked landed with the A/B numbers + the mad2 Narc-profile-under-LRM-label receipt find; drift audit steering row updated (authored cap [T1] over the port-shape blend [T3]); field guide section 8 swarm entry rewritten for the next iteration (found-and-wired, with the two-racks-two-personalities not-a-bug warning).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016bw71WVsccjwW7uKRg4aWD

context/open-questions.md

@@ -1363,8 +1363,12 @@ EXPERT-only; the player+0x260 WRITER has never been read).  Settled/confirmed [w
   thruster[0x3a], rad/s).  The +0x4c candidate hypothesis was WRONG (superseded; +0x4c ->
   thruster[0x3c], role unread).  The binary flies the authored rates; the port's push
   simulator uses uniform PORT-TUNED steering gains (4.0/8.0) [T3] and ignores +0x40.
-  SUNDAY ITEM: consume model+0x40 as the per-round turn gain (the table already caches the
-  record; add the field), bench LRM-vs-Streak turn-radius A/B, expect the 'angry bees'.
+  LANDED SAME DAY (cec6458): the cap is wired (salvo-scoped latch, master + mirror; the
+  blend stays [T3] under an authored-rate cap; BT_TURN_AUTH=0 lever; [projturn] receipts).
+  A/B bench: 43/69 CAP receipts all at 60dps vs ZERO capped-off; blend demand <=~220dps so
+  360 rounds are authentically unconstrained; impacts 25/9 vs 7/4 (no accuracy collapse).
+  DATA FIND: mad2's 'LRM15'-labeled racks resolve 360dps = Narc-profile rounds (the
+  authored-mislabeling lore receipt-confirmed); madcat's LRM15s are true 60dps.
   Same audit, envelope verdict: ttl = lifetime + 10.0 (_DAT_004bf5ac) and kill-plane -1.0
   (_DAT_004bf5b0) are BYTE-FAITHFUL in the port -- #113 closed-faithful, no envelope change.
   [T3 minor ledger]: the binary also silently retires rounds the constraint resolver

docs/WEAPONS_DRIFT_AUDIT.md

@@ -194,7 +194,11 @@ One class, data-differentiated.
 - **Port [T3]:** N-round visual ripple (lead round carries damage; N−1
   damage-0 tracers), ±2.5° deterministic tube spread, steering gains
   4.0/8.0 port-tuned (the old 'MissileTurnGain=4.0' attribution was the
-  proximity fuze), muzzle safety clamp.
+  proximity fuze), muzzle safety clamp.  2026-08-14 (#113 follow-on,
+  cec6458): the AUTHORED per-round turn rate (model+0x40 deg/s, pi/180
+  into the thruster [T1]) now CAPS the blend's per-frame heading change
+  -- 60dps LRM vs 360dps strk/nrk personality is live; the blend under
+  the cap remains the port shape [T3].
 - **Verdict:** KNOWN-DIVERGENCE [T3] (presentation-only; damage economy
   identical).
 

docs/WEAPONS_FIELD_GUIDE.txt

@@ -361,12 +361,16 @@ THE ORIGINAL'S BEHAVIOR
   - Several racks LABELED "LRM" feed Narc-profile rounds; at least
     one "SRM6" feeds Streak rounds. Identical-looking racks homing
     very differently is authentic authored data.
-  - THE SWARM: [Oracle: "The maneuverability of the NARCs and STREAKS
-    often caused an angry swarm of bees effect when the missiles came
-    near a maneuvering target at a high aspect angle."] We are
-    running down whether the original authors a per-round turn rate
-    our port does not read yet -- if your memory holds more detail
-    about WHICH racks swarmed, write it down.
+  - THE SWARM -- FOUND AND WIRED (Sunday's build): Oracle's "angry
+    swarm of bees" led us to the machine's own authored turn rates --
+    60 deg/sec for true LRM rounds, 360 for Streak and Narc rounds --
+    which the original genuinely flew and our port had been smoothing
+    over with one tuned rate for everything. From Sunday's build each
+    round turns at its authored agility: LRM volleys arc heavy and
+    committed, Streaks and Narcs snap. A receipts bonus: some racks
+    LABELED "LRM 15" fire Narc-profile rounds (authored, authentic) --
+    if one LRM rack on your mech snaps while another lumbers, that is
+    the machine, not a bug.
   - THE SRM RANGE PIP -- SETTLED AT 800: the printed manual shows
     450, our content authors 800, and Oracle's note ("agrees with the
     800m range published on the VWE site for v4.10") dates the change