Front end queue growth after failover

Exchange 2003, 2 front end servers and backend cluster in datacenter A. Have built a new 2003 backend cluster in new physical location (datacenter B) and now migrating mailboxes to datacenter B. ESM sees FE's and the 2 virtual servers of each cluster. Pilot group of mailboxes moved and working normally but have noticed that when failing datacenter B active node over to the passive, suddenly the FE at datacenter A queues up email heading to those mailboxes until I fail it back. Any ideas? Thanks!