to attempt scheduling every pending job in priority order. Don't
stop scheduling for a SLURM partition after first job scheduling
failure as many different job parameters are considered: geometry,
rotate, conn-type, node-use.