the longer I’m using Matter over Thread, the more issues I’m running into – and it’s getting quite frustrating. I’d really appreciate some guidance or insights on how to move forward from where I’m currently stuck.
Initially, I started with an OpenThread Border Router and a Sonoff ZBDongle-E 3.0 (EFR32MG21). Everything worked fine — two Eve Door Sensors and a Nuki Smart Lock 5 Pro were operating reliably.
After moving to our new place (a bit larger), my Home Assistant server (HAOS 2025.10 running on a Proxmox machine) now sits in the basement — not exactly the ideal location to cover the entire house with a small USB stick. As a result, I started experiencing random connection losses with the smart lock — surprisingly, not with the Eve Door Sensor right next to it (on the first floor).
My next idea was to add another border router, ideally a PoE-powered device. So I bought an SLZB-06 mg26 and installed it on the first floor.
However, I quickly found out that HA can’t run two OTBR instances at the same time. I tried setting up a standalone OTBR Docker container, but couldn’t get it working via TCP instead of USB (seems this isn’t really intended — and the HA Add-on docs also don’t recommend using a TCP-based RCP).
So, I reversed the setup — used the PoE RCP as HA’s main RCP and planned to run the USB dongle in a separate container later. Unfortunately, this didn’t work either. With the latest SLZB firmware (20250703) and OTBR 2.14, it initially worked fine, but then the OTBR add-on started crashing randomly without the watchdog restarting it. The logs look like this:
[W] P-RadioSpinel-: radio tx timeout [W] P-RadioSpinel-: RCP failure detected [W] P-RadioSpinel-: Trying to recover (1/2) ... [W] P-RadioSpinel-: Error processing result: NoBufs [W] P-RadioSpinel-: Error waiting response: NoBufs ... [C] Platform------: RestoreProperties() at radio_spinel.cpp:2218: Failure
So I went back to the initial setup — the Sonoff Dongle directly connected to Home Assistant, plus two additional Eve Smart Plugs that should, in theory, improve the mesh network.
That setup lasted not even a day. Now I’m stuck with a never-ending loop of:
00:44:44.899 [N] Mle-----------: Different partition (peer:272500625, local:1408724729) 00:44:44.899 [W] Mle-----------: Failed to process Advertisement: LinkMarginLow
I should also mention that there are three Amazon Echos in the house, running their own Thread network — which of course doesn’t integrate into HA’s Thread network. I’d prefer not to use the Amazon devices as the backbone unless that’s the only viable solution.
If anyone has run into similar issues or found a reliable configuration, I’d be very grateful for your insights or recommendations.
Thanks in advance!