What do you mean by 'aggregated links"?
That would be where a "link" between two devices is formed using 2 or more physical links, configured to act as a single virtual link using LACP usually. ("Trunks", "Etherchannel", "802.3ad", various names for it depending on vendor).
I have (rarely) had comms affected by a bug in link aggregation whereby after some random amount of time the aggregated link stops passing *some* traffic. I've had this on Nortel as well as VMWare.
To rule out such a bug, disable all physical links in the aggregated link except for one and monitor for a period
We only lose connection between these two controllers. These two servers are the must busy in the cluster. The ISS server has ~200 users and the MCD controller has all the trunks - both SIP and T1.
Doesn't sound exceptionally busy, however that's all a matter of perspective from the point of view of the switch - that switch could have some kind of per-protocol rate-limiting configured on its in-path interfaces and it is blocking traffic after it reaches the configured limit. Bit of a stretch really, but you never know, those Nexus switches can be a bit odd, I haven't used them for years.