How do you handle data dependency versus task dependency scenarios? For example, if a task completes successfully but the data is incorrect.

Task dependency controls the order in which tasks run; it guarantees that Task B will not start until Task A has finished, regardless of the data produced. Data dependency, on the other hand, ensures that a downstream task only runs when the required data is available and valid. In Airflow, you can enforce task dependencies with the DAG structure (e.g., task_a >> task_b). For data dependencies, you can use sensors (e.g., FileSensor, HttpSensor) or XComs to pass data between tasks. If a task finishes but the data is incorrect, you can add a validation step that raises an exception or triggers a retry. Alternatively, use an ExternalTaskSensor to wait for a downstream DAG that performs data quality checks before allowing the next task to run. This combination of task ordering and data validation ensures both execution order and data integrity.