Jobs provide a means to support distribution of complete flows.
Compared with distributing external tasks, distribution of flows is tricky, since flows contain arbitrary computation. Haskell does not contain a means to distribute arbitrary computation, so additional work is needed by the user to ensure this is possible.
Currently, the only implementation uses Redis to distribute flows, and relies on a map of strings to jobs (of a specific shape - e.g. input and output type) which must be known to all nodes executing the job.