run 3

docs/pt2-state-map.md

@@ -28,6 +28,33 @@ The current evidence suggests the panel behavior is shaped by several layers:
 
 We have good evidence for layers 1-4. Layer 5 is still missing.
 
+## Probable System-Level Flow
+
+At the system level, the best current model is:
+
+1. startup / identity / mode traffic
+2. maintained background status scan
+3. selector/detail pages layered onto that background
+4. operator events flowing back upstream
+5. unknown final condition that makes the panel fully active
+
+This is less like a tiny request/ack protocol and more like a multiplexed
+camera-state bus with a control return path.
+
+Working split:
+
+- **CCU side** probably holds the richer master camera/session state lattice
+- **RCP side** probably behaves like a smart terminal that:
+  - caches a smaller local state
+  - displays received camera/session values
+  - sends button/control events back upstream
+
+That model fits the manuals unusually well:
+
+- LCD/status content such as filter, extender, and self-diagnosis data
+- semi-awake behavior under recurring host traffic
+- broad structured command/slot lattices under several payload surfaces
+
 ## Baseline States
 
 ### S0: Idle / Not Active