I have roundabout 30 Nanoleaf Matter Thread bulbs (downlight) installed.
I have a shared Thread Network between Skyconnect and Apple Homepod.
Turning on and off the light works fast and perfectly.
But if I turn off multiple lights (e.g. 12) at the same time, it takes more than 30seconds (sometimes several minutes) until the state in Home Assistant is turned off. But the lamps are already off in reality.
However, if i stop the OTBR and use Apple as the only BR it works much faster and the status is updated in 2-3 seconds (however this is still quite long). But with this setup I get the same problems if I turn off all 30 lights at the same time. What could be the reason?
Note: wlan is on channel 1, Zigbee on channel 11 and Thread on channel 25.
Second:
if i turn of a light, the switch goes shortly to turned off and then again to turned on and after some time again to turned off.
The problem is bandwidth.
Each device needs to a command sent specifically for it and then it needs to report back that it have changed state.
Matter have a feature called Matter groups, which makes it possible to send a message with a group ID instead and all the devices in that group will then recieve it and act on it.
The messages needed will therefore almost be halfed in number and that lights will act faster.
The reporting back will still be one message per device, but it is still a big improvement.
The problem is that not all device have it implemented and I am not even sure about HA here.
It might have come in the newest Matter Server addon update, but the other devices, like your Nanoleaf ones, need to support it too.
But what I don’t understand: All lamps turn synchronously on and off. There is no latency.
Only the status messages are too late in this case.
If it would be a bandwith problem turning on and off should have the same latency. Or am I wrong?
Furthermore: If it’s a thread bandwidth problem why is it significantly better with Homepod alone than with Homepod and SkyConnect in combination?
TL;DR: Performance of the Thread network is way worse when using two different Thread border routers in the same network and using only one is recommended for now.
You might be hit by Mocelet’s bug link too, but the way the network protocol works makes it way easier for a device that is already transmitting to keep transmitting, simple because it knows when itself stopped and then just have to check for silence and transmit again.
When all the devices need to report back, then the fight starts.
A device will listen for silence and then transmit, but if there is not silence, then it will wait a random time and try again. It is these retries that takes time.