You must call a rate‑limited partner API to enrich records before writing to Silver. The goal is high throughput with bounded concurrency and retries. What Python pattern fits best?

Use asyncio with aiohttp and a bounded semaphore to limit concurrent requests. Wrap each call in an async retry loop (e.g., tenacity or custom logic) that respects the partner’s back‑off policy. The semaphore ensures you never exceed the allowed concurrency, while retries handle transient failures. This pattern scales to thousands of requests per second, keeps memory usage low, and integrates cleanly with the rest of the ETL pipeline.