I used to have my hub reboot weekly. Seemed to keep things humming along nicely. I recently removed it since I had seen a couple mentions of "it shouldn't be needed". After 9 days of uptime things were running VERY slow. Memory was very low.
Is there an official stance on the need for reboots on a schedule?
Is there anything I should be watching for in app/hub/device stats to make things run smoother on their own?
You aren’t the first person to have this issue.
And here’s how I “solve” it.
The official stance is that you do not need to schedule a reboot. Now, there are few exceptions and caveats.
The following conditions may benefit from more frequent reboots:
I just removed it as I don’t use it anymore.
Whenever I've seen this issue it always seems like maker api is involved and is getting too many calls... I certainly could be wrong but when I see smoke I suspect fire..
Despite it not being used by anything, I turned on debug logging, after you brought it up, and it was extremely active.
This is something our engineers are investigating, but the key here is "getting too many calls". We have observed that heavy use of Maker API can cause problems.
Right... I'm just using it as a diagnostics reference not saying that's what it is in this or every case... I try to play process of elimination with a lot of stuff.
Personally, i.e. a non-official recommendation, I would make a note of this memory level, and reboot and observe your hub's behavior. If it consistently starts to slow down at about the same memory point, then I'd consider adding a reboot when it gets within some comfort factor of that number.
Everyone's mix of apps and devices is different, so I always recommend that you understand your hub's mix and how it functions.
If you find yourself in need to restart your desktop often because it's slowing down, then chances are that you will need to restart your hub too 
#GIGO
It depends on the OS and how well they manage memory usage, app segregation and clean up tasks. Eg my Linux and macOS machines happily run for a month or more between reboots.
Windows for PC’s, on the other hand, becomes unstable after about a week!
I have always on PC's and rarely reboot except for the occasional reboots. I have servers running multiple VM's and maybe I'll cycle every few months for updates. If you're running a lot of crapware (weather linux or windows) then there will always be memory/device creep. One of the things I have to constantly explain to clients
Same.
I run Fusion 360 with 20gb+ models on my Mac mini (32gb ram) without issues - I couldn’t do that on my similar specced work supplied windows machine without a weekly reboot. Even when just running multiple office 365 apps and chrome a weekly reboot was needed.
Thankfully I just got given a new 14” MacBook Pro with the M1 Pro SOC - despite only having 16GB of ram, it’s vastly more stable than my Dell was.
Windows Server is a different beast, it’ll run nicely for long periods, likely due to less crap being included.
Back in the day, my Linux servers would get rebooted when patched, my Windows servers would never go beyond 30 days, and some were rebooted every week.
Yeah that’s exactly how I roll, I patch monthly and that’s when the system gets a reboot.
I'm glad to see the more experienced users are doing reboots as well. Things like this always end up making me very paranoid. I end up ruminating over things way too much. I've actually removed some metrics related apps for the simple fact that I had to for ease of mind.
I am not using Maker API.
I THINK my ZWave mesh is in good shape.
I am running a Unifi Protect integration that WAS using a Pi for a bridge but now is able to run just on the hub. Stats wise, it seems to be at the top of my list consistently. Maybe I should be looking at that?
Five minutes after I took this screenshot memory went to 375 range then back up to 383
Kudos to @bobbyD (obviously on behalf of Hubitat) in providing a recommendation from the HE team. An easy minefield to choose to avoid.
Peoples perception of better or poorer performance can cloud judgement of systems like HE, which, personally, I think is unfair. Laptops / Desktops or even rpi's performing very different functions, can produce very different performance outcomes. I think what people need to gauge a HE hub by is their own expectation of what you are asking it to do.... and who you are asking to achieve that outcome. If you are asking HE built-in apps to deliver your outcomes, then it is understandable to question the HE staff on the performance of their hardware and platoform, alongside appropriate details of your setup. If you introduce Community drivers / apps into the mix (particularly anything I have developed 
), then it is fair to expect any conversation around performance to include a critical assessment of your choices of apps and drivers. Thirdly, use of some built-in apps do produce elevate usage of resources on a HE hub, which can be worked through with HE support staff, as @bobbyD has outlined.
BTW - No quams with the OP, just some general comments...
Well put! That brings me back to what I mentioned about my own personal observations. I tend to obsess and worry.
Download the Hubitat app
