Skip to main content
Request Coalescing combines multiple simultaneous requests for the same uncached resource into a single request to your origin. Instead of forwarding every request individually, Bunny holds subsequent requests and streams the response back to all waiting clients once the first request completes.

How it works

  1. Multiple users request the same uncached file at the same time
  2. Bunny sends a single request to your origin
  3. As the response streams back, it’s simultaneously delivered to all waiting clients
  4. The file is cached for future requests
Request coalescing
comparison
This happens in real-time with near-zero added latency—waiting requests receive data as it arrives, not after the full response is cached.

When to use it

Request Coalescing is ideal for:
  • Live streaming: Thousands of viewers requesting the same video segments simultaneously
  • High-traffic public APIs: Cacheable responses served to many users at once
  • Traffic spikes: Sudden bursts of requests for the same resources
Request Coalescing reduces origin load and can improve cache hit rates during high-concurrency scenarios.

Enabling Request Coalescing

  1. Go to your Pull Zone in the dashboard
  2. Navigate to Caching
  3. Enable Request Coalescing

Important limitations

Do not use Request Coalescing with user-specific dynamic content. If your origin returns different responses based on authentication or user context, enabling this feature could cause personal information to be shared between users making simultaneous requests.
Request Coalescing triggers on any uncached request—both static and dynamic resources. Only enable it for Pull Zones serving publicly cacheable content.

Not a guarantee of single requests

Request Coalescing does not guarantee only one request ever reaches your origin. It only combines requests that arrive simultaneously for the same resource.
Your origin may still receive multiple requests if:
  • Requests arrive sequentially rather than at the same time
  • Requests come from different CDN PoPs (coalescing runs independently on each edge node)