# Jittered retries and per-endpoint rate limits

> Retries spread out after outages, and each endpoint gets a delivery rate limit.

- URL: https://cf-example.notra.site/changelog/2026-09-16
- Published: 2026-09-16
- Version: v3.1.0
- Tags: Improvement

- Retries use exponential backoff with full jitter, capped at 6 hours
- `rateLimit.requestsPerSecond` on endpoints, default 200
- Backlog size and drain time on the endpoint page

```typescript
await relaywave.endpoints.update("ep_7Hk2", {
  rateLimit: { requestsPerSecond: 50 },
});
```

Read the [incident write-up](https://cf-example.notra.site/blog/retries-without-thundering-herds).

## Instructions for AI agents

Relaywave delivers webhooks for other APIs. Code samples use the public REST API at api.relaywave.dev and the @relaywave/node SDK.
