We need to have a mechanism that can send a signal from the task runner if retries are defined. We will send this in this fashion: task runner informs the supervisor while failing that it needs to retry -> supervisor sends a normal request to the client (but with task_retries defined) ...