sometimes the 10g scheduler refuses to start jobs any more. I do not know why at the moment, but as a workaround you can just kill -9 the “coordinator job queue (CJQ0)” process from OS level. The process will be named like “ora_cjq0_SID” in a RAC.
PMON does a restart of the coordinator after some sconds, and it works again. May be a hack, but I will tell more as soon as I know it.
Great scheduler post: http://www.lifeaftercoffee.com