Net printers occasionally consuming high CPU usage

High, once in a while (unpredictable) some of our net-printers will
start to eat up almost all CPU cycles.
We realize it because system slows down with any apparent reason.
WRKACTJOB shows one of the writers a using 90+% of CPU... The printout
has already been printed, but for some reason is not SAVEd (all of our
printouts have the SAVE(*YES) parameter) and shows still as *WTR ...
Ending the Spool writer job and restarting the writer usually fixes it,
but the problem is quite annoying since its only after someone complains
about system performance that we realize it.
We are at V4R5, latest Cum.

I've been looking thru the archives to find out if there is an API  we
could use in a batch job inquirying say once every 2 minutes thru the
list of active jobs, to find out if any one job is eating up more than,
say 50%, of CPU so we could automatically kill it and restart it... I
assume there is such a API, but the second questioon is: as I understand
it, WRKACTJOB shows CPU percent as a "mean" figure since last time
statistics were restarted (F10 or F13 ...) so if such an API would
"query" the list of jobs, long running jobs like spool writers would
always give a low figure as a "mean" value, UNLESS statistics were reset
every 2 minutes, maybe by that same API program...

Am I missing something obvious? Can it be done some other way around?
The obvious question is, could someone figure out why any spoolwriter
would eat all CPU and work perfect after END/RESTART ?

TIA
-------------------------
Antonio Fernandez-Vicenti
afvaiv@wanadoo.es