Open issue? ZHA not working properly after upgrade to 2023.5.0 with HUSBZB-1 · Issue #92581 · home-assistant/core · GitHub looks like it. No one from the team has responded AFAICT though.
I installed the latest HUSBZB-1 firmware using the following guide. Re-paired my Aqara door sensors (on 2023.5.2). We’ll see if it makes a difference in ~6 hours.
Worst-case, I think the new HUSBZB-1 firmware is also compatible with Z2M?
I’m having major problems with 2023.5.x as well. Numerous ZHA based zigbee devices going off line. Some Aqara, some Tuya, some Tradfri.
Sometimes I can get them to recover by restarting, sometimes they don’t come back at all, and won’t re-pair to the network.
Often ZHA doesn’t come up properly at all after a restart, and it takes several restarts to get it working again.
The logs often now report that the Zigbee channel (15) is at xxx% utiltised (usually 75% ish). This hasn’t happened before.
have flashed my Sonof Dongle-E with the latest firmware, no improvement.
UPDATE: It’s been 8+ hours now and my Aqara devices seem to be still paired and responding. Updating to the most recent HUSBZB-1 firmware may have done the trick. Running 2023.5.2 still. I’ll let you know if anything wonky happens later today 
I’ve ended up going back to 2023.4.6 which seems stable.
The new warning message about channel utilisation (interference) has been helpful to re-sight my coordinator to a hopefully better position, but I wasn’t suffering any drop outs until 2023.5.x came along, so I guess we’ll wait and see what the devs say about it.
UPDATE:- I reattempted the upgrade to 2023.5.2 and things have been stable for the last few hours. I made 2 changes:-
- Updated my sonoff dongle E firmware, but I am pretty sure I ended up re-flashing the version already on the stick.
- Used a USB cable to move the coordinator further from my wifi router and mini pc running HA.
Not 100% sure if (2) has done anything because I still get warnings that my zigbee channel is at high utilisation (typically 75-80%), but everything seems to running ok at present.
I’ve paired mine multiple times and they keep dying off 
Woke up this morning to find 15 zigbee (ZHA) devices offline. Aqara, Moes, Haozee, Heiman all affected, and except for 2 Aqara in wall switch modules, all are battery powered end devices.
Whatever is happening doesn’t seem to affect things like smart plugs and bulbs.
The location of the devices also seems to be all over the house, and Aqara seems to be the worst affected.
UPDATE:-
I think the dev’s have found the issue:-
4 Likes
The same thing happened to my Phillips remotes on 2023.5.2. None of them would function and the add/remove dance didn’t work either. Rolled back and repaired everything and they work again.
Thank god… Can’t wait for that update to be merged into main!
The devs are suggesting this is affecting ‘older’ coordinators, but I am using a pretty new Sonoff Dongle-E that runs the same chipset as the new SkyConnect, so I think the issue is wider than described.
My dongle does however display as a [HUSBZB-1]-
Appreciate this thread otherwise I would have thought I was going crazy.
My aqara door and motion sensors were all going unavailable, or just not triggering. Guess it started on 2023.05.
I have a HUSBZB-1 on the original firmware with ZHA.
I’ve read plenty of things about aqara being unstable for various reasons, but up until now they’d been rock solid for a couple of years (not using any incompatible repeaters or anything).
I was re-pairing them left and right for the last week, but yesterday I downgraded to 2023.4.6 and it’s been back to normal for 24h now.
1 Like
Still not sure why its affecting my coordinator. According to the comments, it should affect ‘older’ coordinators.
If I download my ZHA debug info, I get this info for my coordinator:
"metadata": {
"ezsp": {
"manufacturer": "",
"board": "",
"version": "6.10.3.0 build 297",
"stack_version": 8,
"can_write_custom_eui64": true
}
Stack version is (I think) ezsp version, so I am at v8, with a pretty recent firmware version.
So why is the bug reported affecting me ?
Has anyone tried upgrading to 2023.5.3 to see if this was fixed? I downgraded to 2023.4.6 and just left it since it was working just fine pre-2023.5.
I eventually had everything working on 2023.5.2 after re-pairing multiple times.
I upgraded to 2023.5.3 this morning and everything is fine so far, no devices were missing after the upgrade. With 2023.5.0, 1, and 2, there were devices that had fallen off right after the upgrade, so I think 5.3 is GTG.
Nope. Woke up this morning and 8 aqara devices are missing. This is really maddening. For something that used to be rock solid to have been suddenly broken by an upgrade is… Discouraging, to say the least.
Hopefully they get a PROPER fix out for this, and soon.
All my ZigBee’s (only 3 devises) went of line with the 2023.5. After updating to 2023.5.3, then resetting all devises, re-installing the ZigBee integration, selecting new network(?) in the setup process, re-paring the devises, all works fine and is stable.
Home Assistant 2023.5.3
Supervisor 2023.04.1
Operating System 10.1
I already went through that process - with 5.0, 5.1, AND 5.2.
Imagine having to do that every single time you upgrade.
Never used to have to do that.
If that’s going to be the way it is going forward, it’s going to cause me to seriously re-think the use of home assistant. These devices are much too important, and there are simply too many of them, for this to be considered even remotely acceptable.
Maybe the " Release 2023.5.3 - May 14" did fix the problem:
- Bump ZHA dependencies #92870:
* bellows to [0.35.3](https://github.com/zigpy/bellows/releases/tag/0.35.3)
This is a bugfix release that fixes some sporadic multi-PAN startup issues, in addition to a critical network formation bug.
Maybe you didn’t read the entire thread…
For those of you that did read it, I just re-paired those devices. AGAIN. We’ll see how long they last this time.
For me, the fix appears to be good. Over a day now and no devices have dropped off.
Still can’t understand why it should have affected my coordinator in the first place, but the fix does appear effective (at least for me).