We often have a combination of small resource rules (a few mins, <10MB RAM) and large resource rules (hours/days and >10GB RAM). It would be helpful to be able to specify the cluster queue and resource requirements in the rule definition. This proposal is similar to Issue
#44 (Allow "local" jobs when submitting to a cluster).
My solution at the moment is to use the --cluster option to specify the largest job's queue and resource requirements, but this results in our 'big job' queue being filled with a large number of tiny jobs that don't need to be there. One downside with this proposal is that it would make the Snakefile less portable across cluster systems.