I've seen this ONCE on our box. We had to reboot - since our problem was with
the HPSYSJQ which you can't delete.
I'm assuming some system table is getting corrupted - I noticed that once corrupted,
the value remained stable (but wrong) until the reboot.
Here's our implementation of job queues.
We have put all the permanent job streams into a background job queue - and set the
system job queue limit to 1. There are a couple of other job queues which we use
to fast track (or slow track) individual (and non-critical) jobs.
CURRENT LIMITS ARE 20 FOR JOBS
JOBQ LIMIT EXEC TOTAL
HPSYSJQ 1 0 0
BACKJQ 13 10 10
SLOWJQ 1 0 0
FASTJQ 1 0 0
The beauty of this approach is that we retain single threading for our overnight
job streams, and the failure of any background job does not impact on that single
threading.
When we had our problem with the job queue's counts, I simply set the limit on
job q with the problem to a maximum value, but ensured single threading of the all jobs
by setting the global job limit to be just ONE greater than the number of
permanent jobs.