Once the job is running you are executing service application on your compute node.Moving job between nodes would mean to run you service application on different compute node which in turn means to close your application (or wait until it ends) and re-run
it on new compute node.
I would suggest to wait until your service applications end (or some of your taks assigned to the job end) and re-execute job with different settings (nodegroup,numcores etc) or just cancel running job and re-execute it with new parameters.I don't think
you can change job's resources durnig runtime however you have shrink/grow option available in scheduler settings.
Daniel Drypczewski