Zigbee issue with Homey Pro 2026 migrated from Homey Pro (Early 2023) using USB Restore

Questions & Help
124

Hello everyone,

I’m back after some time away.

Since the Zigbee reset, everything related to Zigbee has been working well for me.

My problem is that I can no longer add Matter devices and have to connect to the user first. If nothing happens soon, I will probably send the HP 26 back and switch to HA. It’s a shame and not what I wanted, but Athom obviously isn’t interested in the problems.

How did I come to this conclusion?

I requested support as described above and am posting my email and their response here:

Jan 13, 2026, 14:16 GMT+1

Hello Arno,
Adding it wasn’t the problem, but rather that the sensor was no longer accessible after a while. But it was still on the network.So there was no error message with the red triangle, just no response from the sensor.
Nevertheless, I tried it, created a back-up and reset the Zigbee network:
I have already documented this in the community case, and I am including it here:

11 January 2026
Briefdescription of the current situation and starting point:

I had aHomey Mini – Pro, but I quickly reached the limits of its memory with mydevices, so I decided to get a Homey Pro 2026. To avoid starting from scratch,I made a backup of the Homey Pro Mini. I then imported this when the bigbrother arrived.

Allbattery-powered devices in my case, Aqara window/contact sensors that wereconnected after importing the backup, were only usable for a period of X.

This meansthat pairing is possible and visible in Homey (no red triangle), but there areno more notifications when doors/windows are opened. Repairs, restarting Homeyand re-importing backups did not lead to any improvements or solutions.

1. Backupcreated locally.

2. Resetthe ZigBee network using the developer tool.

  • In my case, I did this twice because the first reset caused the ZigBee channel to change to “11”, which was too close to my Wi-Fi channel.
  • So I reset it again and got channel 20.

image792×336 28 KB

3. AllZigBee devices were marked with a red triangle.

4. Selectedthe device → “Device not available” → you can click directly on “Try torepair”.

5. Put thedevice into pairing mode.

6. Thedevice was quickly reconnected.

This is thestatus at 1:20 p.m.:

I tookadvantage of the short break to report to you that 20 powered devices arecurrently connected.

This ispurely theoretical, so that the Zigbee network may stabilise in my mind.

Now I willcontinue and connect the battery-powered devices. In my case:

Aqarawindow/door contacts, motion detectors, water sensors and temperature sensors.

I will getback to you with an update once this has been done.

________
11 January 2026 - 4:30p.m. → All devices are connected.
The watersensor/leak sensor caused some problems → but gave up after a few attempts.

Ireconnected two window sensors after about an hour and changed the battery.

Otherwise,everything looks good – time will tell what the actual result will be.

I’m stillhaving problems with the two Matter devices from Ikea – I haven’t been able toconnect them yet.

I willprovide a long-term/24-hour update tomorrow.

______
11 January 2026 – 4 p.m. → Current status – 2 sensors had to be repaired again.
Some of thecontact sensors need to be opened and closed 3–4 times before they respond.

I am notparticularly satisfied so far, or rather disappointed → another 6 hours of workand even the most basic things are not working. I don’t even need to thinkabout flows etc. as long as the sensors are so unreliable. Reminder: thesepreviously worked 100% reliably with HomeBridge under Apple Home.

I don’tknow whether a ZigBee network like this still needs to “find its feet” …
Status - 12 January 2026
As thingsstand, it looks good; the Zigbee network seems to have stabilized.
6:00 p.m.

The Matterdevices have unfortunately stopped working again.
End of documentation

Finalquestion: this cannot be the solution, because what happens if I have a crashand need to restore a backup? That would be a vicious circle, wouldn’t it?

Kind regards,

Dennis

________________________________________________________________________________________

Answer:

Hi Dennis,

Thanks for your response and detailed feedback

Just keep me posted

I look forward to hear from you.
Thanks in advance.

______________________________________________________________________________________

That’s ridiculous. The existing problems are not addressed, and I assume that the email was not read in its entirety. We, as end users, do the analysis for Athom, and then we don’t even get a proper response. How rude!

My response will now be summarised: if and when can we expect a fix? Otherwise, I was a Homey user for the shortest amount of time possible.

125

I have replied and hope that we can get some clarity for everyone.

At least, that’s what I hope. We’ll see. Have a nice Sunday, everyone.

Reply:

Hi Arno,

Sorry, but I’m going to be honest here.

You can’t be serious with that answer!

I’ve carried out a very detailed analysis here and documented everything, as have many other users in the case mentioned.

This is not our job, but Athom’s.

Finally, I still have problems with Matter and when restoring a backup, which you don’t even address.

The HP26 costs 400€ and is completely faulty.

If it crashes and a backup is restored, it has to be completely reconfigured or the devices have to be reconnected → (after the ZigBee reset).

Is that supposed to be the solution?

This makes a backup completely obsolete.

What is the planned course of action here?

I would like a statement as to whether the problems mentioned here have been addressed in any way and whether there is a planned software/hardware fix for them.

If nothing is planned, I reserve the right to return my Homey by 25 January 2026.

I am speaking primarily for myself here, but judging by the forum, you will lose many customers with the current service and support.

1 Like
126

@SunBeech but is is sure after a reset everything is working fine ?

Only thing is that o need to repair the devices flows and everything after repair is working again then ?

Did you tried it already ?

And after the reset can I add the frient smoke detector ?

Now I got this message

image0
image01991×2868 125 KB

127

See the post of @Jonathan_Draper:

As the Zigbee devices are not removed from Homey after the Zigbee reset (just unavailable), you don’t have to repair your flows. You do have to repair all your Zigbee devices. Note: repair, not re-pair (so via the Maintenance > Try to repair option).

I haven’t tried this myself, but you can find multiple (recent) reports on this forum on this behavior, like the one linked above. Some other posts:

I don’t think anyone can provide you with assurances that you specific device will work. You’ll just have to try it yourself. But if your device worked on your HP23, then it should also work on the HP26.

128

The problem is that restoring doesn’t work, or only for a short time!

I returned my HP 2026. My 20203 restarts occasionally due to memory issues, but that’s just the way it is until a solution is found.

129

Yes, that is what this topic is about.

However my prior post was about the work-around: restore a back-up followed by a Zigbee reset and repair of Zigbee devices. That work-around seems to work. But is not ideal if you have a large Zigbee-network.

130

Hey Alexandre… did you manage to find a solution?

Im at the stage of refunding:

  • all my matter plugs that aren’t working
  • Zigbee plugs that are doing the same as you
  • and the 2026 which just has so so many issues…..

Any help welcome

Daniel

131

I reached out to Athom again regarding the issue (thanks @Doekse for clarifying this matter, and feel free to correct me if I am wrong).

The TLDR is that this is indeed a known issue and something that most likely cannot be fully fixed. The larger the Zigbee network, the more likely a Zigbee reset is required to get it operational again.

The longer explanation: the element that is causing the issue is the Zigbee routing table Homey maintains of your Zigbee network and connected devices. Apparently this table is stored in the Zigbee chip itself and cannot be accessed by the Homey OS. As such this routing table is not included in the backup nor restore of your Homey.

The consequence is that, after the restore, your new Homey first needs to rebuild that table and remap your Zigbee network. The problem is that while Homey tries to do that, Zigbee devices might flood the Zigbee network trying to find the old Homey/the old Zigbee coordinator. This excessive communication prevents Homey from repairing existing, adding new Zigbee devices and rebuilding the Zigbee network. The larger your Zigbee network the more prevalent this issue is and the less likely that Homey will succeed in rebuilding the network. For smaller networks, you might be lucky that Homey is able to rebuild its Zigbee network.

Though Athom indicated that they might be able to enhance the restore/rebuild process, it would not be a definitive solution for larger Zigbee networks. The only remaining option is to reset your Zigbee network on the new Homey and repair all devices (via the Maintenance > Try to repair option).

HELP! NEW Homey2026: restored from backup
132

Perhaps they should take a look at how Z2M deals with this, because with it, a migration between similar chipsets can be done without a reset or re-pairing :man_shrugging:t3:

133

Not for me when switching to my Sonoff Dongle Max MG24, exact same issues as when switching to the Homey Pro (2026) :sweat_smile:

Some Silvercrest Tuya Zigbee plugs were spamming te network with 4 messages per second.

To elaborate on @SunBeech post: The Silicon Labs Zigbee/Thread chipsets used in both Homey Pro (Early 2023) and Homey Pro (2026) store their routing tables directly on the chip itself in non-volatile* memory. This data cannot be exported or restored. This isn’t a limitation on Homeys side, Silicon Labs just doesn’t offer that functionality. As a result, when a backup is restored or a coordinator is replaced, the Zigbee coordinator has to rediscover and rebuild the entire network from scratch. Sometimes this works, but especially in larger networks this process can be slow, unstable, or fail altogether.

Put differently, it’s like handing a complete stranger a valid passport and dropping them into a room full of people who all recognize that passport and assume they already know him. On paper, his identity checks out. In reality, he doesn’t know anyone there. He now has to rebuild all relationships from scratch… while everyone in the room is talking at the same time and shouting over each other, assuming he already knows who they are.

In practice, this means the Zigbee coordinator must reconstruct the entire network topology (rediscovering devices, routes, and relationships) while traffic is already flowing. If that traffic is too noisy or unstable, the coordinator may never converge on a stable network.

In my case, I had several Silvercrest (Tuya-based) Zigbee plugs acting as routers on both my Homey and Zigbee2MQTT networks. After switching to Homey Pro (2026) / a Sonoff Dongle Max (MG24), those plugs appeared to assume their previous parent or routes were still valid. When they failed to properly reattach to the new coordinator, they repeatedly initiated route discovery and rejoin attempts, effectively flooding the network with routing traffic. This prevented other devices from successfully rebuilding the mesh, causing the network to remain unstable.

After removing all the these from power the network could breath again.

So in short; there are hundreds of things that could influence such a back-up and restore which makes it really hard to troubleshoot.

1 Like
134

Thnx for this explanation. That is why modded homeys pro 2023 do not have this issue?

135

They use the same Zigbee chipset so the routing table the chip has build isn’t “lost” so to say

1 Like
136

Non-volatile, I presume :smiley:

Never mind, I think I understand the process now. The controller will build up a new routing table with every reboot, but that’s not a problem.

Does Homey restore the IEEE address of the old Homey to the new one? Because it sounds like it doesn’t, in which case there’s a big chance that you will run into the problems that you describe. The Z2M documentation explicitly states that it should be copied to a new dongle if you want to perform a migration that requires the least amount of re-pairing.

137

Actually, I compared the System Information - shown in the Zigbee section of Homey Developer Tools - on my HP23 before the backup and my HP26 after the restore as input for my original support request. All settings are the same except the IEEE Address.

1 Like
138

That would be a good explanation of the problem then.

139

Oh no :see_no_evil_monkey: I’ve sent my 2023HP back!!!

140

Thank you for the explanation. Appreciate your update. I have 3 follow-ups on solving the issue.

I have 44 Zigbee devices, so I assume this is considered a large network. From you descriptions of the issue, would it be possible to power down all Zigbee devices and then power them up one by one?

Or even better, how can I see what devices are spamming the network?

I also have 7 Matter over Thread devices from IKEA. 3 powered and 4 battery. All battery devices are dead as well. Since they share radio, will they start working again when Zigbee is stable?

141

That’s indeed what I did, but halfway through I decided to just reset the network since I was accessing all devices anyway.

A reset sounds scarier than it actually is: all devices remain known to Homey, you just need to re-pair them one by one after the reset. Same goes for your Matter over Thread devices indeed.

For me it was the perfect opportunity to kick the Tuya Zigbee devices out of the house and replace them with some discounted IKEA alternatives :wink:

142

Good one! That might be causing some issues then. I do know swapping the IEEE doesn’t come without issues though, is does also mess with the security keys which can cause even more issue.

143

An IEEE address is like a MAC address, right? Then it would need to be cloned, but that means the old Homey won’t work anymore (since 2 networks cannot share the same IEEE address). Or am I wrong?