run 35

docs/discovery-notes.md

@@ -7509,6 +7509,183 @@ Reasoning:
   - set-once then maintain-later
   - beacon first, page second
 
+### 2026-05-13 HE32 Results
+
+Front-panel result:
+
+- No HE32 startup/handshake sequence visibly woke the panel.
+- No useful LCD, lamp, or readout change was observed.
+
+Serially, though, the ladder was still informative: the panel did not become
+"live", but several of the startup-shaped sequences diverted into alternate
+one-shot response families instead of opening the expected page streams.
+
+#### HE32a: Discovery-first, then `90 -> E8`, then maintained `7A`
+
+Log:
+
+- `captures/rcp-he32-discovery-then-e8-maintain.txt`
+
+Result:
+
+- `A0` produced no structured reply in this run.
+- `B0` produced the known readable block:
+  - `07 80 6C 40 30 C1`
+- `B5` then produced no new stage.
+- The later `90 -> E8 -> 7A` portion stayed heartbeat-compatible only.
+
+Read:
+
+- Discovery-first ordering did **not** turn later page maintenance into a live
+  session.
+- It mostly behaved like "consume one discovery/readable block, then collapse."
+
+#### HE32b: `A0 -> 90 -> E8`, then heartbeat-only maintenance
+
+Log:
+
+- `captures/rcp-he32-a0-90-e8-heartbeat-tail.txt`
+
+Result:
+
+- `A0 -> 90` produced a new family:
+  - `07 80 64 40 30 C9`
+- After that, `E8`, `7A`, and the heartbeat tail stayed heartbeat-compatible.
+
+Read:
+
+- This is a useful new lead.
+- `A0 -> 90` does not look like a wake path, but it does look like a distinct
+  startup/opening branch rather than random failure.
+
+#### HE32c: `A0 -> 90 -> E8`, then `7A`-only maintenance
+
+Log:
+
+- `captures/rcp-he32-a0-90-e8-7a-tail.txt`
+
+Result:
+
+- This reproduced the same startup branch more strongly:
+  - `07 80 64 40 30 C9`
+- Repeated `7A 50 26` after that did not open a new stage.
+
+Read:
+
+- `64 40 30` now looks real, not a one-off accident.
+- It is associated with the `A0 -> 90` stacked-opener idea, not with a live
+  maintained page stream.
+
+#### HE32d: `AF -> 90 -> E8`, then maintained `FA`
+
+Log:
+
+- `captures/rcp-he32-af-90-e8-fa-tail.txt`
+
+Result:
+
+- `AF -> 90` also diverted into the same family:
+  - `07 80 64 40 30 C9`
+- The later `E8` and repeated `FA` host feed stayed heartbeat-compatible only.
+
+Read:
+
+- `64 40 30` is not unique to `A0 -> 90`; `AF -> 90` can also land there.
+- That makes it look more like a startup/opening class than a branch-local oddity.
+
+#### HE32e: `A0 -> AF -> EC`, then maintained `7B`
+
+Log:
+
+- `captures/rcp-he32-a0-af-ec-7b-tail.txt`
+
+Result:
+
+- `A0 -> AF` produced another distinct family:
+  - `07 80 0D 04 AB 7F`
+- `EC` and the repeated `7B` host feed after that stayed heartbeat-compatible.
+
+Read:
+
+- This is another useful startup-shaped diversion:
+  `A0 -> AF` appears to open a discovery/status-like family of its own.
+- But again, it did **not** enable the later page/value stream.
+
+#### HE32f: Repeated `90` beacon, then single `E8` page
+
+Log:
+
+- `captures/rcp-he32-90-beacon-then-e8.txt`
+
+Result:
+
+- This capture contains two runs.
+- Run 1:
+  - repeated `90` beacon produced `07 80 64 40 30 C9`
+- Run 2:
+  - repeated `90` beacon produced `07 80 E4 40 30 49`
+- In both runs, the later `E8 -> 7A` portion stayed heartbeat-compatible only.
+
+Read:
+
+- Repeated `90` by itself is not waking the panel, but it is clearly not
+  meaningless.
+- It can steer the panel into at least two startup/readable families:
+  - `64 40 30`
+  - `E4 40 30`
+
+#### HE32g: Repeated `AF` beacon, then single `EC` page
+
+Log:
+
+- `captures/rcp-he32-af-beacon-then-ec.txt`
+
+Result:
+
+- Repeated `AF` beacon produced:
+  - `07 80 0D 04 EB 3F`
+- The later `EC -> 7B` portion stayed heartbeat-compatible only.
+
+Read:
+
+- Repeated `AF` also looks meaningful as a startup beacon, but again only in
+  the sense of opening a one-shot family, not in waking the panel into a live
+  session.
+
+### HE32 Overall Read
+
+This ladder did not wake the panel, but it did sharpen the handshake model:
+
+- the panel is still **not** waking on our best "declare then maintain" startup
+  sequences
+- but stacked openers and beacon-like traffic are clearly meaningful
+
+New startup-shaped families observed:
+
+- `A0 -> 90` or `AF -> 90` -> `07 80 64 40 30 C9`
+- repeated `90` beacon -> `07 80 64 40 30 C9` or `07 80 E4 40 30 49`
+- `A0 -> AF` -> `07 80 0D 04 AB 7F`
+- repeated `AF` beacon -> `07 80 0D 04 EB 3F`
+
+Best current interpretation:
+
+- `90` and `AF` are increasingly looking like **mode/class beacons or startup
+  selectors**, not just ordinary page openers
+- but the missing wake condition is still something beyond those beacon effects
+- once one of these startup families is opened, the later page/value stream
+  still collapses to heartbeat instead of becoming active
+
+So HE32 rules out another important family of guesses:
+
+- the panel does not appear to wake just because we got the startup **order**
+  closer to reality
+
+It may still require:
+
+- a different class of maintained traffic after those beacon stages
+- a missing hardware/state signal on another wire
+- or a specific startup family we have touched indirectly but not yet maintained
+
 ### 2026-05-13 CAM POWER Context Retests
 
 Goal: