rig: pin the rig-nodes package set 0.4.0 at its release tag (LAB-338)

rig/README.md

@@ -10,7 +10,7 @@ This directory is its own pixi workspace. The root `pixi.toml` and `fictrac-brid
 untouched, so bridge 3.4 (`pixi run bridge` at the repo root) stays installed side by side.
 
 > **Since 2026-10-26 (LAB-338):** the shared part of this graph — `rig_interfaces`, the device
-> adapters and simulators (FicTrac, Kinefly, BIAS, ScanImage), the MCAP recorder/indexer, the CDR
+> adapters and simulators (FicTrac, Kinefly, BIAS), the MCAP recorder/indexer, the CDR
 > decoder, the generic Foxglove rewriter, the supervision primitives, the Zenoh configs and the
 > benches — lives in **[reiserlab/rig-nodes](https://github.com/reiserlab/rig-nodes)** and is
 > installed here as the pinned `rig_nodes` (+ opt-in `rig_interfaces`) packages from that repo's

rig/WINDOWS-TEST.md

@@ -161,7 +161,7 @@ whether the router keeps running after Ctrl-C (`tasklist | findstr rmw_zenohd`).
 pixi run test -- --keep
 ```
 
-Expected: `54 / 54 checks passed` (57 / 57 with `-e typed`; 59 / 62 before the ScanImage connector left on 2026-10-28) in about a minute and a line `artifacts kept in …`. Keep that
+Expected: `60 / 60 checks passed` (63 / 63 with `-e typed`; measured on the Mac 2026-10-28 with rig-nodes 0.4.0 — the six host-monitor checks of #282 raised the count, the five ScanImage-connector checks left with the connector the same day) in about a minute and a line `artifacts kept in …`. Keep that
 directory (`summary.md`, `crash-results.json`, the converted export, `runlogs\*.bag`, `proc-logs\`).
 Windows-specific things it exercises for the first time: child processes in their own group,
 `CTRL_BREAK_EVENT` as the stop signal, `TerminateProcess` for the mid-run kills (the FicTrac adapter

rig/docs/okey-cokey-foxglove-dashboard-plan.md

@@ -54,7 +54,7 @@ afterwards).
 | BIAS health | `/bias/status` at 5 Hz (`bias_node.py`): reachable, capturing, logging, frameCount, framesPerSec, rtt_ms; on Atlas10 also frameIdGaps, cameraSkipped, loggerSkipped, writerSkipped, logQueueSize. `viz_node` turns these into `/viz/bias/*` and a `bias` entry in `/diagnostics`. `foxglove/arena-bias.json` shows them. | Mostly there. Missing: a rate from the frameCount slope (independent of BIAS's own estimate) and the drop counters on a plot. |
 | Arena health | `viz_node`'s `arena` entry is fed only by `/arena/frame`, which only FicTrac lines produce. | **Always "never seen" on this rig.** |
 | What the arena is doing | Every Studio message (trialParams `a` echoes, runner status events with `conditionName` / `params` / `durationSec`, controller telemetry `cc`/`cf`/`cs` row batches) reaches `bridge_node`. | It is **written to the bag only** (`/arena/browser`, `/arena/event` via `rec.write_raw`), never published, so nothing live can subscribe. The `a` hex is truncated to 8 bytes (no gain/duration). |
-| Expected display image | `tools/patfile.py` reads `.pat`; `tools/tiff_run_to_mcap.py` already renders `/arena/display` as a `foxglove.CompressedImage` (200×40, upscaled, labeled RECONSTRUCTED). | Offline only, and wired to the 2P color-sweep file names and DO1 timing. Monochrome. |
+| Expected display image | `tools/patfile.py` reads `.pat`; the 2P tiff-run converter (`tiff_run_to_mcap.py`, private since 2026-10-28 in scanimage-2026-bergamo `2p-rig-nodes/`) renders `/arena/display` as a `foxglove.CompressedImage` (200×40, upscaled, labeled RECONSTRUCTED). | Offline only, in the private repo, and wired to the 2P color-sweep file names and DO1 timing. Monochrome. |
 | Camera image | Nothing. BIAS's HTTP API has no frame grab (only status, timestamps, frame count, video file). | **No live frames anywhere.** After the run, the `.ufmf` plus the per-frame stamps on `/bias/frames` (recorded at stopRecording) are enough to rebuild a timed image stream. |
 
 ## Design