some snmp agents do not support interface index persistence. At the time those agents will be restarted, interfaces of a device will get random / different interface indicies.
If such an interface with no corresponding MIB table entry has a connection modelled in Spectrum, a stale interface alarm will be triggered. After some time, the stale interface will disappear from the interface table and there will be a notice of n unresolved connection(s) in the upper right of the interface table:
This idea is to implement some behaviour to get detailed information for the unresolved connections. What are the target devices and interfaces of those connections?
A workaround is to rightclick and resolve the connection (Maybe incorrectly, as you see only the target device, not the interface) to see it in the topology view, maybe delete and recreate it. It would be better to eg. have this notice clickable and get a list of devices and interfaces for which the connection cannot be longer resolved. This would help to troubleshoot and remodel those unresolved connections correctly on the first try.