After updating NSPanel Pro to firmware 4.3.0, the Zigbee subsystem becomes unstable. Zigbee devices randomly switch to offline state and then reconnect after a few seconds. This happens without any identifiable trigger and affects different devices at different times.
After several such events, the Zigbee stack appears to stop functioning entirely:
all Zigbee devices show as offline,
the list of synchronized devices becomes empty,
MQTT bridge stops publishing Zigbee data,
In the eWeLink app, under NSPanel Pro → Home Assistant/MQTT, every action fails, indicating that the internal MQTT bridge stops working once the Zigbee stack crashes.,
only a full reboot of NSPanel Pro restores functionality.
This strongly suggests a regression or memory/queue issue in the Zigbee stack introduced in firmware 4.3.0.
Impact:
Zigbee devices become unreliable,
automations fail unpredictably,
MQTT integration becomes unusable until reboot,
NSPanel Pro cannot be used as a stable Zigbee gateway.
Expected behavior:
Zigbee network should remain stable, devices should not randomly disconnect, and MQTT bridge should continue to operate without requiring manual reboot.
Suggested direction:
Investigate Zigbee stack stability in firmware 4.3.0, especially memory leaks, task starvation, or watchdog resets affecting the Zigbee coordinator and MQTT bridge.
I have sent several reports from the NSPanel Pro: 1001ea2549.
I apologize for the issue. We have also identified this problem and are currently working on a resolution. A new version addressing this issue is expected to be deployed to your device tomorrow.
Hi, a dedicated version 4.3.2 has been deployed for your device. You can now proceed with the upgrade. If you encounter any issues, please feel free to contact me.
@Milk After updating the NSPanel Pro firmware (4.3.2) that was intended to fix Zigbee stack issues, the device has once again started publishing raw internal data structures to MQTT, which should not be sent in any production build. Home Assistant, using MQTT Discovery, interprets these payloads as valid sensors and automatically creates dozens of incorrect entities.
This behavior existed in older firmware versions, was partially fixed in the previous release, and has now returned - this appears to be a clear regression.
Examples of incorrect payloads published by NSPanel Pro
The device is sending, among others:
Apologies for this issue. This issue has been fixed in v4.3.2. Please send us your device ID via private message, and we will push this targeted update to you.
What does the fix cover
Inhave spent many (too many) hours trying to install the SNZB 02DR2
App is V5.22
NS panel is V4.3
SNZB is 1.0.3 ( after a very difficult upgrade process)
1 The graph in EWElink shows no hourly figures
2 The device shows as not supported in Ewelink Web
3. There is only one sensor in HA to display in that app - no temperature
I also experience that the Zigbee devices randomly disconnect. When this happens, I restart it, and then it works fine again for a while. The automations are not reliable, so I would also try version 4.3.2.
My device ID: 10026564ee
After several days on firmware 4.3.2, I can confirm that the Zigbee stability issues have been resolved. This also results in stable MQTT server operation - stable enough for reliable, production use.
However, on 2 or 3 occasions the NSPanel Pro unexpectedly stopped working in a way that suggests the watchdog triggered a reboot, after which the device got stuck on the CUBE startup screen. The screensaver with the clock remained visible, and only touching the screen revealed the problem. After “waking up” the device, CUBE loaded normally and everything returned to proper operation, including Zigbee.
Today I also submitted feedback from the NSPanel Pro [1001ea2549], right after functionality was restored.