Queues are not limited I believe, but probably should be to prevent the docker storage filling up. Will need to look into what options we have when the queue gets to the limit.

Apply backpressure or delete old messages? There are tradeoffs. Backpressure might cause the upstream system to start deleting messages, which is arguably worse. As long as we've saved them to disk, we can replay them.

Also we should be monitoring the queues so we get plenty of warning.