Job priorities

We don't use Dataset queues, only separate queues for the different kind of tasks (see known defers)

This introduces an expected problem as Procrastinate will process tasks first in, first out: A small job triggered from the UI might have to wait 2 weeks until many (long running) jobs for another dataset are finished.

To avoid this, we assign a random priority to each job during deferring. UI-Triggered (User CRUD) jobs get a higher random priority.

The randomness range can be configured for each known task queue via it's settings:

OPENALEPH_INDEX_DEFER_MIN_PRIORITY=75

• Defer settings reference
• Read more about priorities in procrastinate docs