Latency can be a key consideration. Async style code can make it a lot easier to fire off a set of requests in parallel and wait for all the responses to come b

Latency can be a key consideration. Async style code can make it a lot easier to fire off a set of requests in parallel and wait for all the responses to come back (e.g. with https://docs.hhvm.com/hack/asynchronous-operations/concurrent https://docs.hhvm.com/hack/asynchronous-operations/concurren...), rather than the serial request waterfall a basic thread-per-request model will push you towards.