Currently the publication job client queue grows unbounded and is cleared at restart. When the monitor returns we should persist the details into the database and remove the operation from the active queue. Also implies the status query needs to have multiple phases.