The Job Manager accepts a some options that control its willingness to run jobs in parallel.
Of particular use are the options --parallel-limit and --distribution-parallel-limit. Optimal values for these options depend on your model and what analysis you are running on it, as well as on the properties of the node's hardware.