Quarry currently will report that something is "running" forever, which makes it seem like there's a stuck queue job or that the database is allowing it to run forever when anything happens to a celery worker. This is quite misleading. There must be some way to detect that the worker is gone/dead/broken and return something better ("killed" would work, but a more explicit status would be better).
This would promote better bug reporting when things are down and shorten time to recover. It would also help users understand what's happening in general if they ran a query that fills the system RAM or something that isn't immediately fixable.