Some connections for a collapsed SSE stream dropped early: is there a limit?

We’ve started using SSE request collapsing in production and we’re seeing a weird mix of client connection times. Some stay connected as expected for many seconds, up to our 5 minute limit, but many are connecting, getting a 200 response and disconnecting in under a second (in fact, just a few hundreds of microseconds because the request is a HIT, so this rules out any of the connection timeouts I was previously dealing with)
There are several thousand clients that would resolve to the same cache key, so I’m wondering if we’re hitting some limits and the connections over the limit are getting the cached result and dropped.
All of the clients i’m looking at now are using python requests library under our custom code so they have a lot in common.

Any ideas? Thank you

Hey @jkebinger, no product-level limits would be causing this, which leads me to believe that this may be a bug or be linked to a specific nuance of our application. Given that, I’ll convert this to a support ticket so we can dig in a little deeper.

To close this out, this was due to documented behavior when the backend server drops the fastly connection before the cache expiry. We wrote about it more at Our Experience using Fastly's Streaming Miss feature For Server Sent Events (SSE) | Prefab

1 Like