An additional question: now that the director is running again (after a reboot, actually), I notice that the catalog says the failed job 14976 is still running, while the director says it isn’t (and actually doesn’t even include it in the ‘status’ output, neither as running nor as terminated).

Does that faulty catalog state need to be fixed, or will it go away when the maximum run time is expired?

If necessary, how can it be fixed?  Google doesn’t seem to be able to find the answer to that right away.

I restarted the job to see of the problem comes back, and the director does list the new instance as job 14977.