I have a use case where I want to make request to another endpoint in the same compute service. So I’ve configured a “self_backend” host for this purpose. Atm I just get 503. Is this in principle possible?
We do have a configuration option which can be set to allow ‘loop’ requests like this, but it is not often used because the risk of harm to the platform is quite large since a misconfigured service could endlessly loop requests to itself. If there is any other way to accomplish your goal that would be a better choice 
I’m current using Request::set_after_send. But the problem is that it can’t be used with Request::send_async, only with Request::send. So my idea is an internal endpoint that returns the object I’d otherwise be putting in the cache with set_after_send - that would ensure normal cache semantics for the derived object in the cache.
If the sdk allowed set_after_set with send_async then I wouldn’t need to deal with this really. The docs imply without saying explicitly that this might change: Request in fastly - Rust .. note the “does not currently support”
That makes sense… to get more details on future plans for the SDK to support the cache-control callbacks in async mode you’d need to open a support ticket or ask your account manager to get the information.