My issue was older very chatty Z-Wave devices that I made worse by setting the them to poll every second (I was warned and ignored it, so that's totally on me). They were indeed where my problem was. Same as you, I long suspected them, but I liked the use cases with this high frequency polling rate so much, that I was just living with the slowdowns they were causing by doing nightly reboots.
There have been changes to the platform that have improved things, but (and this is an absolute guess here) anecdotally, it seems like earlier versions of the platform were more tolerant to the problem devices. When I was running 2.1.3.x (for around a year) the issue really didn't present itself fast enough for me to draw a conclusion to what the issue was. But when I finally decided I was going to move forward and get up to the latest hub platform version, I was getting the same slowdowns after 24 hours or so that many others have been experiencing. I will say that a total lockup rarely occurred, but things would become so unresponsive, that it seemed like it was locked up.
After moving my Aeon v1 HEMs to a second HE hub, my main hub is now working like a champ. No more slowdowns. I know it was the chatty devices, since the issue of slowdown has now moved to the second hub, which only has those HEMs on it, and when that was previously my test hub, it never slowed down. So I've solved the issue by simply auto rebooting that hub at 4am each day. That's a very cost effective solution for me, since purchasing two brand new Aeon v5 HEMs would be $300 CAD and there is no guarantee that having them refreshing at that rate would not just duplicate the issue on newer hardware.
