14 months!? OK, you really need a battery tester. No wonder you're getting weird drops after 20 seconds. I'm fairly certain you're running into issues due to low battery. The readings from the driver may not be accurate. You need to test them or simply replace.
Here's the type I use. It's not as accurate as a ZTS, but it's good enough for regular lithium cells and affordable.
Thats an erroneous assumption on your end.
Why are you doing assumptions without knowing for example the battery reads?
In fact these sensors use a 3V CR2450 battery, I did a test (you can see that I also swap batteries between 2 devices) the readings on the Aqara Sensor (brand new) was 3.02 and the reading on the original motion sensor was 2.64v so both still had plenty juice level.
So let's not assume things. We can ask people to test the battery levels for example but let's not say that is for certain this or that without that data to corroborate. I have used a multimeter.
But thanks for those links it might came in end as they seem easier to use than a multimeter.
Besides a bad battery on one or two devices would not make them ALL disconnect like it was the case on the second time that devices dropped out.
I know English is not your first language, but your responses come off a bit rude. I'm only trying to offer suggestions to help. That's all I want to do for anyone here. Sorry if you took my ribbing as an attack on you. Not at all the intention.
I don't see how your stating that your batteries are 14 months old and then my repeating those exact words here are an assumption of anything. You are using a method that isn't even recommended by even the author of the driver to test accurate battery levels. Your reading of 2.64v is low and likely to causing the issue. Likely, does not mean absolutely is.
My writing I'm "fairly certain" about something, means I'm not 100% certain. So you can ease up a bit and not cast aspersions on people. I never wrote that I was certain. I wrote fairly certain. It means one is almost certain, but does have some doubt. That doubt could possibly be proven true or false by actually removing the battery, putting in a proper battery tester and finding out if it's indeed weak. I'm guessing it is too weak for proper performance. Most people's experience, my own included, with not only Xiaomi but also other Zigbee devices is that they do not perform properly when the batteries are low. Low is going to be subjective, based on the device, not one's interpretation of what an acceptable battery level is. They're all going to act differently at different minimum battery levels. I don't know what the level is that's acceptable for Xiaomi or any other for that matter, but personally if I have trouble with a device, replacing the battery is the very first thing I'm going to do. Even if I measure it with a battery tester, I would still replace it. A measured value in most battery testers is completely different than a measured value under load. So I'm not claiming that's a perfect guide and the absolute definitive solution. The ZTS testers I believe do put the battery under load to test, but that's neither here, nor there.
To me, and this is just my opinion, I'm not claiming it is fact, but rather a guess based on experience, that the low battery level is almost certainly the cause of any Zigbee device misbehaving. Don't believe me? I don't need you to. Ask around. But please don't tell me that you don't need to change batteries, can't figure out why your devices are quickly dropping off the network and that obviously it's something wrong with the Hubitat Zigbee implementation. No sale here sir!
Good explanation.
Was not trying to be rude to you. Was just making pass the point across that I had tested batteries.
If my method of testing batteries is incorrect I don't know but was the method for this type of batteries I was told to use on the electronic store I purchased the multimeter.
If a battery that according to them is more than half full is "low" for Xiaomi devices i dont know. But is odd if thats the case.
Again I'm not saying that the fault is on the platform or on the devices.
It become one of my pet peeves people do judgments based on empiric evidence (experience) when i saw one of my employees generate more that $1.2m extra cost as they trusted their experience.
So when someone's makes statements like the one you reference I just go ballistic. Nothing related to you.
Give me technical explanations, corroborated facts, etc
As such is not I don't believe you... It's just I need more than empirical evidence.
P.s I didn't took your response as an attack.
I have no issues with what either of you are saying. Except...
The batteries are not 'half full'. They are almost dead.
Most devices have a minimum operating voltage - and it isn't 0v. For standard 3v devices, minimum voltage is typically considered something between 2.7v and 2.3v. Some devices are more sensitive than others, so there is no single correct number for this.
There are many articles out there on this, feel free to look it up. Some conservative guides say never use a standard battery that is <10% below face voltage - which for these would be 2.7v.
I usually consider 2.7v time to replace, but not an emergency, as most devices will work down to 2.5v.
Also keep in mind that a bettery will test different on a multimeter than it will under load. 2.7v on a multimeter could be 2.5v in operation. That's why if you are that far down on the curve, it is better to just replace them.
Thanks for those numbers.
I have ordered 100 new batteries and will replace them.
And last, but not least, what I said above only applies to traditional/standard batteries. Rechargeable and lithium batteries have different voltage/capacity curves, and much different voltage guidelines.
I have so many battery operated things now I stock up on batteries when I see sales. They are never on sale when I need them. I'm still waiting on my first to die (or get close to).
Just about any CR2450 battery that measures less than 2.8V open circuit (without a load) is pretty much dead, at least based on what I've seen using them with SmartThings multisensors (haven't had to replace one in a Xiaomi yet). To get an idea of their remaining capacity using a multimeter (which has a very high impedance, effectively an open circuit) you need to use a load resistor.
This post may be of interest : batteries - CR2450 Battery at 2.95 Volts - Electrical Engineering Stack Exchange
TLDR: depending on application, CR2450 measuring 2.95V open circuit could have about 10% or 50% capacity left
another option is to modify the devices from batteries to dedicated hard wired power using usb cables or usb adapters.
The Xiaomi devices are very easy to modify if you are handy with a soldering iron.
Thanks for that article interesting read.
Question: the article refers 'Based on Voc=2.95V or -0.05V from 3.0V, I estimate it is at 90% SoC" but you mention 10 or 50% why the difference? Load?
No plugs where I want the sensors hence the need for batteries. Being a rental I can't change the wiring as well. 
I prefer batteries over cords everywhere. Especially if the batteries last a year. I can take 30 min once a year and replace them all at once.
Each to their own.
So I paired 3 of my 10 Aquara temp/humidity sensors yesterday. So far all 3 check in every hour as designed - which is cool.
I think I'm going to have to take a calibrated thermometer to each of them to figure out a temp offset though, as they all read different temps by 2-4 DegF when they should be basically the same.
Humidity is also about 7% different across them, but humidity readings are often +/- 5%, so that might be as good as that will get.
Yes, that would be an estimated remaining capacity depending on the ESR of the load in the range of the example.
Got that at the begining with ST. There were 1 or 2 with different readings. One case I really had different readings because of the surface I had attached the sensor. I attached to an aluminium window frame. So the conductive of the material was affecting the readings.
The second case I had 2 of them on the same location and were giving me a few differences. At the time I couldn't troubleshoot as I was going on holidays so left them in a shelve. When I comeback they were giving nearly the same temp (with a 0.2 difference).
I currently have Netatmo health coach plus the temp sensor in the same room and readings are pretty in sync
Question for anyone wit the temp/humidity sensors...
If wall mounted via the adhesive back, how are you changing the batteries?
I hopefully won't have to change any batteries soon, but was thinking about that this morning...
Yes this is how I do it.
P.s this will not work on the new Vibration Sensor or the water leak
Interesting. Do you have the old round, or the new square ones? Maybe it doesn't matter.
I'll have to get brave and try one. 
The round ones.
But the adhesive will hint if it's possible
Download the Hubitat app
