Failed ingest jobs block system from dispatching other jobs

The ingest service creates non-dispatchable jobs in the service registry when adding a new track. This allows to distribute the ingest load across several ingest nodes.

In case of an exception, the job is not always set to "FAILED". This is an issue because running jobs add to the node/system load. The stale ingest jobs pretend that there's still load on the system, even though there is no process running anymore.