I am investigating this as well.
It is a new site, or office, in a new location, so we have may network issues as well. The Interaction Tracker log will show an alert was offered, alerted, and ACD interaction not answered by agent A. (agent did not see an alert)
Then call is offered/alert to another agent B.
The two entries are in the same HH:mm:ss even thought the no answer time is 30 sec. Agent A status is changed to Agent not answering status.
I'll look in TsServer logs. We use remote numbers, so the SIPEngine may not show what I am looking for.
My theory is that the CIC did not get a software response back from Agent A's client? So it changes them to not answering so they are required to change to available, as a fail-safe for a client side network or application failure?