Chris Yes, i am logging in as Super User also, other side same user's connection(Superuser) i am trying to kill.
Can you give a try, by picking the IP's of the processes which are not getting killed from pg_stat_activity(client_addr column) and make REJECT entries in pg_hba.conf file as below:
host all all <client_addr IP> reject
and reload the cluster using pg_ctl -D $PGDATA reload