If a job run on container agent takes longer than 10 minutes to spin up, the pod is killed and the job fails. Allowing for a configurable task timeout would support jobs that take longer to spin up, reducing failures due to orchestration.