Rate limits

Test mode has a ceiling, so a load test against it slows down your own test requests and nobody else’s:

• 25 requests a second per test key, and 100 a second per account over all its test keys. Each has a second’s worth to spend at once, so a short burst passes and a steady stream above the rate does not.
• Live mode has no such limit.

A request over either is refused before anything runs:

Response · 429 Too Many Requests

HTTP/1.1 429 Too Many Requests
Retry-After: 1

{"error": {"type": "rate_limit_error", "code": "rate_limit", "message": "Too many requests in test mode: at most 25 per second per key."}}

• The status is 429, the type rate_limit_error and the code rate_limit. The message names the limit you went over, the key’s or the account’s.
• Retry-After is the number of seconds to wait before you send it again.
• Nothing ran, so the request is safe to retry: send a create again with the same Idempotency-Key.

Handling it

• Wait at least Retry-After seconds, then retry. Back off further if the next retry is refused too.
• Spread a load test over time, not over keys: the account’s limit counts every test key of the account together.
• In CI, run tests that create payments one after another, or a few at a time, rather than all at once.