The streaming healthy clients are wrong

[dabaosun]

I'm using 3 nodes and server received their data. But in the UI , the streaming healthy clients are always 1 or 2 clients and person count is 1 or 2 .

And, even if power off the 3 nodes, the Data source always shows 'Real hardware connected'.

[ruvnet]

Thanks for the clear report. Two separate issues here:

1. "Streaming healthy clients" shows 1-2 instead of 3

This is the same per-tick node reporting from #282. The server processes one UDP frame per tick cycle, so the node count reflects whichever ESP32 packet arrived in that particular tick — not the total connected nodes. All 3 nodes are cycling through (you'd see different node IDs in the debug logs). This is cosmetic.

2. "Real hardware connected" persists after powering off nodes — BUG (now fixed)

This was a real bug. The server sets source = "esp32" when the first frame arrives but never reverts it when frames stop. So if you power off all 3 nodes, the UI keeps showing "Real hardware connected" indefinitely.

Fix (branch fix/esp32-source-timeout):

• Added a last_esp32_frame timestamp to track when the last UDP packet was received
• Added a 5-second timeout — if no ESP32 frame arrives within 5s, the source changes to "esp32:offline" in all status/health API responses
• The system health endpoint now shows "degraded" instead of "healthy" when ESP32 goes offline

This means the UI will show the hardware as disconnected ~5 seconds after you power off the nodes. When you power them back on, it immediately switches back to "esp32".

PR incoming shortly.

[ruvnet]

Fixed in PR #300 (merged). The server now detects ESP32 offline after 5 seconds of no frames and reports "esp32:offline" + "degraded" hardware status. Pull latest main to pick up the fix.