Now that we can generate 429 responses for rate-limited requests, a further enhancement would be to add a Retry-After` header that indicates when to retry. The api-client library could use that information to (optionally) delay (sleep) before retrying instead of throwing an exception (and having people home-roll their own retry policies)