The script pulls data from an API, and the API fails on about one call in 50. What does that mean, and how would you handle it?

An error rate of 2 % (one failure per 50 calls) indicates that the API is unreliable or that the client is not handling transient errors properly. In a production pipeline, even a small failure rate can accumulate to many lost records. To mitigate this, I would implement a retry strategy with exponential back‑off and jitter to avoid thundering herd problems. I would also add a circuit breaker that trips after a threshold of consecutive failures, temporarily halting requests until the service recovers. Additionally, I would log each failure with request metadata and aggregate the metrics in a monitoring system to detect patterns. If the API provides idempotency keys, I would use them to safely retry without duplicating data. Finally, I would set up alerts for the failure rate so that the operations team can investigate the upstream issue. This combination of retries, circuit breaking, and observability ensures that the pipeline remains robust despite occasional API hiccups.