The nodemgr on either the primary or secondary should be kicking out event log messages that might help indicate why failover is being triggered.
To just get things going you might try bouncing the primary server while the secondary is active. After the primary comes back up, look through it to make sure things started and are running smoothly. Then give the Failback another shot.
If you want to do more troubleshooting first, you could try disabling Automatic Failover. Then give Failback a shot and watch the event logs.
-Eric