Throttling

Throttling is intentionally limiting how fast a client’s requests are processed — delaying, queuing, or rejecting requests beyond an allowed rate — to protect a service from overload or to enforce fair use and quotas. It is the enforcement mechanism behind rate limiting.

Worked example: an API returns HTTP 429 (Too Many Requests) with a Retry-After header once a client exceeds 100 requests per minute, so one heavy user cannot starve the rest. Gotcha: throttle at the right layer and granularity — per-user or per-key, not just globally — and return a clear 429 with backoff guidance so well-behaved clients can slow down; silent dropping or a generic 500 makes clients retry harder and worsen the overload.