Say VM with MIP-IP is on DPN connected to controller-1, when MIP is learnt , it populates the cache alivenessCache . When VM migrates to another DPN, which is connected to say controller-2, the same alivenessCache will not have data w.r.t the monitoring session created by controller-1. Hence incase of cleanup , it causes issues in stopping ArpMonitoring. The cache is simple local cache which doesnt work incase of cluster.