For jobs with a job load of >1.0, there is the risk that they never get processed. This is due to the fact, that each host has a maximum load and only jobs which fit into this max load get processed.
Jobs with a bigger load than the most powerful host in the system need to be dispatched anyway.
This is a design feature so that jobs of certain type (say, with a given encoding profile) will only be processed by certain workers. These (presumably expensive) workers can be brought up and down as demand requires.