If the connected interface is going into an error-disabled state due to loop-detection during a power-outage, most likely some other piece of equipment is behaving strangely reflecting the loop-detection frames (PDU for Layer-2 where "Packet" is the PDU for Layer-3) are being reflected back at the ICX unit.
Specifically, the way loop-detection works is that the same switch that generates the loop-detection will recognize it should it come back into another Interface. This normally happens when there is a loop in the network topology.
Usually when you see a route change notification in the logging, it is because the ARP table reflects the next-hop IP available via a MAC address that is reachable via a different interface.
Does the device you are reaching go through something else and have two (2) connections to your Rucks ICX devie?
You will need to explain more about your network topology to get better support.
I can assure you the loop-detection works extremely well in every implementation we have tried. Its only downside we have found is if you have multiple units they pass it through transparently... i.e. you run loop-detection on a VLAN, if that VLAN crosses a trunk (i.e. a link between switches carrying multiple VLANs) that if someone loops something (i.e. connecting an Avaya VoIP phone on two interfaces that go to different closets or switches), the loop-detection frame will get picked up and passed through the other devices which do not recognize it as theirs, and it gets back to the switch that made the loop-detection frame only on a trunk or fiber link.
All that said, what about just not making loops? or using spanning-tree, which is excellent at mitigating loops?