Unequal loadbalancing for a UDP VIP

We've an UDP VIP configured with udp-datagram profile enabled. We're observing a strange behaviour the way LTM routes the packets to the backend server.  There are 5 servers in the pool and most of connections are routed to one server due to this behaviour the resource on that perticular server getting exhausted. 

I am not sure when we've enabled on data-gram profile. I am assuming the some of the old connections still in the connection table and being handled without data-gram. I tried clear the connection table using below tmm script and did not have any success. Please let us know if there are any other way to clear the connection table. 

tmsh show sys connection ss-server-addr <server-ip> | grep tmm | awk '{split($1, client, ":"); print "tmsh delete sys conn cs-client-addr " client[1] " cs-client-port " client[2] " cs-server-addr <VIP> cs-server-port 1813"}' | sh