What's the actual, Oracle related, problem, if any?
For ps command output perhaps some AIX forum is better suited, or
$ man ps
and read about the use of brackets.
The processes seem to have spawned within a few seconds (from the ps command)... Is this temporary? What happened a second or two earlier? Two listener processes were started Mar 28. Could you bounce these?
Are you using Shared Server configuration? If you have longer than very short queries, that would show your symptoms, though only through the listener. But you are sure BEQ connections also show the problem?
Also, take a 15 minute statspack when the problem is happening, perhaps there will be a clue.
Are you seeing any swapping on the OS side? I've seen cases where just a couple more users add enough PGA usage to send the box into a tailspin.