we're using clusters that require different slurm configs, hence the
need for fine grained control.