diff --git a/engine/MUNGA_L4/L4CTRL.cpp b/engine/MUNGA_L4/L4CTRL.cpp
index 5540b95..491f82e 100644
--- a/engine/MUNGA_L4/L4CTRL.cpp
+++ b/engine/MUNGA_L4/L4CTRL.cpp
@@ -2631,16 +2631,21 @@ void
 		//----------------------------------------
 		buttonActivateModeMask[rio_event.Data.Unit] = mode_mask;
 
-		// THE ENG-PAGE EJECT KEY [T3 intent-completion, #118]: side keys
-		// 0x0B/0x23/0x03 are the eng-page EJECT position (engEject lamp 0x85
-		// = kBTEngBankTop-4; the streamed weapon pages route the same key to
-		// EjectAmmo).  Eject was UNFINISHED in the 4.10 binary (the armed
-		// mode never rose), so the generator-page press has no authored
-		// route; complete the intent here IN THE DRAIN (game thread --
-		// mode_mask in hand; a first cut queried the mode manager from the
-		// window-click thread and crashed on the unbound TU global): while
-		// PANIC is armed AND that bank shows an ENG page, forward the press
-		// as a pilot-keypad key -- the binary's one armed eject binding.
+		// THE ENG-PAGE EJECT KEY [T3 PORT CONVENIENCE -- a marked DEVIATION,
+		// #118]: side keys 0x0B/0x23/0x03 are the eng-page EJECT position
+		// (engEject lamp 0x85 = kBTEngBankTop-4).  IN THE BINARY these soft
+		// keys NEVER pilot-eject: every page routes them to a real authored
+		// function (weapon pages: EjectAmmo 0xB; myomer pages: seek toggle 9;
+		// the streamed .CTL dump has no 0x200000 records), and the flashing
+		// engEject lamp is the INVITE -- the 1995 press was the pilot KEYPAD
+		// / PANIC key beside it.  The pod eject chain itself was LIVE in 4.10
+		// (master perf FUN_004a9b5c+0x10 evaluates per frame; an earlier
+		// "unfinished" note here was wrong).  This hook adds the glass-layer
+		// convenience the operator asked for: while PANIC is armed AND that
+		// bank shows an ENG page, the flashing key ALSO forwards as a
+		// pilot-keypad press.  Runs in the drain (game thread -- mode_mask
+		// in hand; a first cut queried the mode manager from the window
+		// thread and crashed on the unbound TU global).
 		{
 			const int  a = rio_event.Data.Unit;
 			unsigned   engBits =
diff --git a/game/reconstructed/mech.cpp b/game/reconstructed/mech.cpp
index c724b7f..41b930c 100644
--- a/game/reconstructed/mech.cpp
+++ b/game/reconstructed/mech.cpp
@@ -572,10 +572,13 @@ int
 	extern int BTPlayerExperienceSimLive(void *owner_mech);		// btplayer.cpp (+0x25c;
 	int noviceSim = (BTPlayerExperienceSimLive(this) == 0);		//  NULL-player reads LIVE)
 
-	// COOLANT-CLAUSE HYSTERESIS [T3 intent-completion, 2026-08-03]: the
-	// binary's clause is a plain `< 0.05` compare -- but this evaluator was
-	// DEAD CODE in shipped 4.10 (never called), so the boundary flap was
-	// never exercised.  Live, a bank fraction hovering at 0.05 under fire
+	// COOLANT-CLAUSE HYSTERESIS [T3 smoothing deviation, 2026-08-03]: the
+	// binary's clause is a plain `< 0.05` compare, evaluated per frame from
+	// the master performance (FUN_004a9b5c+0x10 -- the arcade DID run this;
+	// an earlier "dead code" note here was wrong, corrected same day: the
+	// absolute-pointer scan missed the E8-relative call).  A bank fraction
+	// hovering exactly at the threshold is a degenerate oscillation the pod
+	// plausibly never sat in (leaks drain THROUGH the boundary); live under
 	// flips the panic-arm mode EVERY FRAME (observed: hundreds of ON/off
 	// pairs), churning the mode mask, the lamp, and the arm audio cue into a
 	// ~10 Hz stutter tick.  The engine's own idiom for exactly this is the
diff --git a/scratchpad/SESSION_HANDOFF_20260803.md b/scratchpad/SESSION_HANDOFF_20260803.md
index f84b939..75a679b 100644
--- a/scratchpad/SESSION_HANDOFF_20260803.md
+++ b/scratchpad/SESSION_HANDOFF_20260803.md
@@ -75,8 +75,17 @@ open-questions (i860 DIV-CARD firmware ask).
   = ejectPermitted; review mode = GLOBAL DAT_004fd550).
 - EjectPilot @0049f854 confirmed via the image message table ('EjectPilot' id 0x19);
   DuckRequest = id 0x1a @0049fa00 (mech+0x398 latch, NO reader — vestigial).
-- KEY FINDING: **eject was UNFINISHED in shipped 4.10** — the evaluator had zero
-  callers/pointers until the master-perf call was restored; the arcade never armed.
+- ⚠ CORRECTED (same night): an earlier "eject was UNFINISHED/dead code in 4.10"
+  claim was WRONG — the absolute-pointer scan missed the E8-relative call; the
+  master perf (FUN_004a9b5c+0x10) evaluates per frame, so the ARCADE's eject was
+  LIVE. Only the PORT was missing the call (our PerformAndWatch recon predated
+  knowing the export-hole fn's contents). Swept from code comments same night.
+- THE PRESS MODEL [settled]: pilot eject = the PILOT KEYPAD bank while armed
+  (+ the PANIC key, which reports through that matrix). The MFD soft keys NEVER
+  pilot-eject in the binary — every page routes them to authored functions
+  (weapon eng: EjectAmmo; myomer eng: seek toggle) — the flashing engEject cell
+  is the INVITE LAMP. Our page-gated eng-key eject is a MARKED PORT CONVENIENCE
+  (deviation), operator-requested; drop it if strictness wins.
 - Inputs, all live: Panic button (0x3D→pilot-keypad wire), NUMPAD = pilot keypad
   (CONTROLS.MAP `keypad pilot` bindings), Backspace/pad action, and the eng-page
   EJECT key (0x0B/0x23/0x03) page-gated in the RIO drain [T3 intent-completion].