Handling traffic spikes

StreamlinePay, your payments API runs behind Amazon API Gateway with a usage plan configured for a steady rate limit of 100 requests per second and a burst limit of 200. During a flash sale, a partner integration sends a short, sharp spike of traffic that briefly climbs above the steady rate before settling back down.

Use the simulator to build intuition: drag the incoming requests per second slider upward and watch the accepted count (green) and the throttled (429 Too Many Requests) count (red) respond. Notice the band of traffic just above the rate limit where requests are still accepted, and where 429s begin.

Question: Which mechanism absorbs that short spike above the steady rate limit before any requests are throttled?