Five Minutes of Thirst

An API limits each key with a token bucket of capacity 60 that refills at 5 tokens per second, one token per request. A client sends nothing for five minutes, then fires 200 requests as fast as it can. What happens?

A

All 200 are rejected, because a burst that far above the refill rate empties the bucket before it is checked

B

The first 60 go through immediately, and the rest are admitted at roughly 5 per second as tokens refill

C

Only 5 go through in that first second, because the refill rate caps what any single second can serve

D

All 200 go through, because five idle minutes accumulated 1500 tokens for the client to spend