Rate limits explained: per-minute, per-hour and per-day
A verification API layers several rate limits. Understanding how they interact stops you designing a job that works in testing and throttles in production.
On this page
The layers
Typically a per-minute burst limit, a per-hour sustained limit, a per-day cap, and a per-IP limit. A batch job can be under the per-minute limit and still hit the per-hour ceiling halfway through. Plan for the tightest limit that applies over your job duration.
Design for headroom
Target 60 to 70 percent of the sustained limit, not 100 percent. Retries, other jobs, and real-time traffic all share the budget. A job that just fits will fail the day anything else runs.
Read the retry-after
When you do get a 429, the response tells you how long to wait. Honour it exactly with a little jitter. Retrying immediately just extends the throttle.