The Oban setup of a dandelion app: what queues exist, the schedule, and how a job left by a dead node is given back.

{Oban, Dandelion.Queue.config(otp_app: :my_app, repo: MyApp.Repo, crontab: MyApp.Cron.schedule())}

• default and ordered queues, 10 jobs at once each, on every node. ordered is for Dandelion.Queue.Ordered workers.
• crontab: [{cron expression, worker}], run by one node only (Oban's leader), once per tick for the whole cluster.
• A job left executing by a node that died is given back after rescue_after — five minutes by default. A job that genuinely runs longer may then run twice: keep jobs short, or raise it.
• Finished jobs are kept a week.

config :my_app, Oban, … overrides any of it (tests use testing: :manual). Options here: :otp_app and :repo (required), :crontab (default none).