In case of failure put the task back at the end of the queue

Currently if a playbook fails, the worker gets stuck, we should probably avoid this situation and instead, put the task back at the end of the queue and keep on processing messages.
Otherwise, we can easily end up in a situation where all the workers are stuck by tasks that are at the front of the queue, so re-starting the workers will not help going further.