Fix the Failure Fast
The goal is not heroic debugging. The goal is to know whether the problem is camera, edge, browser, device, or configuration, and recover quickly.
First Question: What Is Actually Broken?
Use the dashboard and current diagnostics to separate source failures from viewer failures.
Camera or Feed
If the room has no camera participant or no video track, start with the local camera source and the edge publisher path.
Edge Runtime
If the edge heartbeat is stale or clip backlog is growing, the local edge machine is the first place to look.
Device Runtime
If the rig is offline, stale, or not reporting presence and telemetry, the hardware-side runtime or network path is likely the problem.
Browser Session
If mobile works and one long-lived Chromium session does not, you may be dealing with a browser-session-specific failure rather than a dead feed.
Common Failure Patterns
These are the practical failure modes we already know about.
Viewer Says the Feed Is Dead, But Mobile Works
Treat this as a browser-runtime problem first. We have already seen long-lived Chromium sessions wedge while the actual room and feed stay healthy.
Clips Stop Appearing
Check edge heartbeat, clip backlog, and last clip worker result. If clips are pending forever, the edge clip worker is the likely failure point.
Control Feels Bad
Look at device freshness, RSSI, ack/result latency, and local Wi-Fi conditions before blaming the backend first.
Rig Is Powered but Offline
Power does not prove network health. If presence and telemetry are stale, the device is effectively disconnected even if the LED is on.
What to Check in the Dashboard
The current platform already exposes useful bring-up and troubleshooting data. Use it.
Rig Health
Check last presence, telemetry freshness, command latency, and signal quality before you start rewriting firmware.
Stream Debug
Use room state, camera track count, restart command, viewer diagnostics, and edge runtime state to isolate video-path failures.
Motion Diagnostics
Use sweep and nudge tests to separate hardware, firmware, and browser-control issues from each other.
Schedule and Overrides
If a node seems unavailable, confirm that it is not simply following schedule rules, date overrides, or a forced closed state.
