@gopher.ny fingers crossed you have some ideas.
I was using the standard Google Fiber hub as my primary router.
I'd hard coded the IP address for the hub on the hub because the GF router didn't seem to do well at using DHCP to assign reserved addresses.
To get a better router (at least I thought!) and wifi6, I went with an Asus RT-AX89X.
While it seems generally better (and it's DHCP actually seems to work, at least initially, my main hub has dropped off the network a couple or so times in the past 2 days.
For whatever reason, it at least thinks it loses it's network connection (I was able to get into the diag port because I had gone back and hard coded the hub's IP on the hub, which let me do a reboot).
Whatever happened, the hub largely seemed to stop working altogether once it got unhappy (it wasn't seeming to run rules or trigger events or anything).
The Asus router is less than a week old.
Using DHCP or static IP, it connects perfectly and runs for a while.
I did see logs on the router complaining that the hub's MAC was improperly trying to set an IP address (e.g., "xx:xx:xx:xx:xx:xx not mesh client, can't update it's ip").
The main hub is connected directly into the new router.
Couple of things to try, in that order:
Settings/Networking now has a speed section. This makes a difference in connection stability on some routers.
If that selection makes no difference, there's an endpoint that makes hub try reconnecting using DHCP when connection drops. Each reconnect slowly eats into OS memory, but it works pretty well otherwise. If hub has to reconnect couple times a day, the memory difference is negligible. If router drops connection every few minutes, it adds up, and scheduled reboot is a good idea, along with looking at router settings. Here are the (self-explanatory) endpoints:
In upcoming 2.2.8, getting 10 or more reconnects an hour will generate an alert and a location event.
After waking up to it being totally nonresponsive, I rebooted. Now, I see that nearly every device command is throwing a Java error. While I see incoming events, essentially ALL actions attempted by the hub trigger the error and fail to execute. Odd.
I'm did a soft reset/restore and that seemed to fix things, but it was rather unexpected.
Now, only a few hours later, I got a low memory alert.
So far, I've only set the speed to fixed.
And, removed a few basic rules and created 3 simple RM ones (to control a new ceiling fan switch I just got).
This may be a big D'Oh. Moment. We'll see.
It appears the default router's IP changed. Not sure why things got messed up with DHCP enabled--but I've changed the "hard coded" IP info now. We'll see what happens.
-> Not much. Still seeing unhappiness. I put the router back to the .1.1 address.
Sorta related and discussed here: Chromecast Integration (beta) Stop Working - #13 by rob9
Going to the new hub caused all my Chromecast things to get new IP addresses--and that busted them as well. Setting them via reserved IPs back to the original IP address fixed them.
I actually wonder if the Chromecast app may have been part of the problem--if it couldn't find the devices, it may have consumed resources and/or otherwise messed things up. Guess we'll see. I got the default gateway set back and Chromecast fixed. We'll see if it starts behaving.
Could be, Chromecast implementation attempts to reestablish active connection when they drop, with multiple retries.