Your question is Idempotent Resilient ETL Design. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How do you design ETL/ELT processes that are idempotent and resilient to failure?
Describe a practical design using tools such as Apache Airflow, Apache Spark, object storage, and a warehouse. Explain how you handle retries, partial loads, duplicate records, late-arriving data, schema changes, backfills, checkpoints, transactional writes, and recovery after orchestration or infrastructure failures. Include the data-quality checks, audit metadata, monitoring, and alerting you would implement.
