Airflow import errors suppressed after modifying HelperDcmWeb.py via codeserver; errors only appear after restarting Airflow pods

Steps to Reproduce

1. In codeserver, open HelperDcmWeb.py
2. Comment out or delete all code in the file; save the changes.
3. Trigger any Airflow DAG that imports HelperDcmWeb (e.g. service-process-incoming-dcm).
4. Observe the task logs and Airflow web-UI.

---

Expected Behaviour

• The DAG (and all tasks that rely on HelperDcmWeb) should fail immediately with an explicit Python ImportError, clearly visible in the Airflow web-UI task logs and in task status.

---

Actual Behaviour

• No import error is surfaced in the Airflow web-UI; tasks appear to fail generically without a stack trace.
• Error message for the failed task:

*** Could not read served logs: Invalid URL 'http://:8793/log/dag_id=pancreas_segmentation/run_id=pancreas_segmentation-250507143649969663/task_id=get-input-data/attempt=3.log': No host supplied

• The actual error is only shown in the logs of airflow-scheduler.
• After executing a restart of all Airflow pods, the expected ImportError traceback becomes visible in the web-ui.

---

Work-around

• Manually restart all Airflow pods after editing any shared Python helper module to force module reload and surface errors.