You need to verify that the following attributes are the same on the model (across all the port and application models): agent_port, community_name, polling_interval, trycount, timeout.
If all of the attribute values are the same across the ports and applications, AND the device has redundancy enabled, then you must install a patch to resolve this issue.This issue has been fixed in H12 and above by swbug023261. The release note shows the following:
Instead of being suppressed, a critical alarm is asserted on ports or application models, which are not polling on their own, when the device is lost.
Non-self polling ports or application models will be suppressed when the device is lost.