So I was able to find out the culprits tying up my 4510's cpu with some help from the forum using
debug platform packet all receive buffer
sh platform cpu packet buffered
and it appears to be redhat hosts using bonded interfaces and and our netapps san using bonded interfaces. I think the bonded interfaces are presenting the same mac address coming from two different ports and this is gobbling up the cpu. For the moment I've unplugged one of the links in the bond, but am looking for a solution on the cisco side to this while the application owners of the netapp and redhat boxes look for a solution on their side. Any one have experience with a similar issue?
Update on this, changing the bond configuration from the default round robin to Adaptive Load Balancing appears to have corrected the problem on the Redhat servers.