yes. It doesn’t have to be powered. I like powered however because I feel a separate power supply for the zwave radio helps. I don’t know if this will work, but it is the easiest to try.
I’m not an expert on all the reasons the controller will reject commands. However, one major candidate is that it is overwhelmed (not defective). If you look at the timing in the snippet the binding is set to wait 250 ms before resending the command, but after 3 times it gives up. The binding looks good to me. That leads back to the ideas by others regarding zombie nodes. Also, if I picked this up correctly from your file you have polling at 2 hours. I have polling for powered devices at 10 days (the max) and for battery devices 1 day (just to get a daily battery reading). Increasing the polling interval probably won’t solve the problem, but it could help reduce traffic. I also do not use the command poll, but that is controversial. Between auto-update and visual confirmation (yes, the light is on!) I feel I’m covered. A great post on Zwave issues is here. IMO -I’d try to get the current situation better, before going the docker route
Lastly I have no experience with docker but agree it does seem popular. I have two OH/zwave instances; an Rpi4 with a zooz zstick on a powered hub (for radio power purity) and a rpi3b with an aeotec zstick directly plugged in (old model that will not run directly off the Rpi4- found out the hard way). Both very stable. I also have a zniffer to see what is going on with the controller, but that is for another day.