Batch queue processing is handled much differently in the >=8.9 releases. It is no longer required to set them up via the RunOneWorld/EndOneWorld.sh scripts (for UNIX) or JDE.INI for NT, and the last time I checked the binary which fired off in Xe is no longer present in the 8.9x Tools Releases. You now setup queues via P986130 and they are updated in realtime by the kernel. You can actually add a job queue or increase the number of threads on a multi-threaded queue without taking services down on your Enterprise server.
Queues are now interupt driven and are managed by the Queue Kernel (Type Def 14). There is a setting you can put in the NETWORK QUEUE SETTINGS section called QKOnIdle. See below, but it does not seem to have any relationship to the amount of time before a batch job is picked up by the queue kernel. It seems to be geared more towards allowing the process wait time instead of chewing up resources searching for new jobs.
Someone please correct me if you have more information or insight in this area. In my experience it is now more a factor of the speed of the server itself which determines the job submission delay.
[NETWORK QUEUE SETTINGS]
QKOnIdle=60
