Worked on reliability and resource management for the KubernetesJobOperator in the astronomer/airflow repository, focusing on resolving monitoring pod cleanup leaks. Addressed a persistent bug by implementing robust cleanup logic using Python, Airflow, and Kubernetes, ensuring that monitoring pods are reliably deleted after job execution and that job-level exceptions are preserved for accurate troubleshooting. Enhanced the operator’s cleanup flow by wrapping execution in try/finally blocks and extending on_kill behavior to cover foreground pods. Expanded unit testing with comprehensive mocks and autospecs to validate edge cases, and updated documentation and changelogs to clarify the improved cleanup contract and operational reliability.