The order may be an illusion, if they are running in parallel the order of the queries in ids may not match the order of the queries in the analysis tool

If you can try them truly sequentially one after the other if the queries all take about the same time, then you know it is something about the running in parallel (contention)

A connection pool in the aws library could make this behavior

Would time waiting on the connection pool show up in x-ray?

the x-ray segment is measured locally, I'd assume it includes a period wider than the network traffic

Ah

(like forming the request, signing it...)

besides map + future isn't always a great pattern. In this case the futures aren't launched until p/all ostensibly forces them

@ghadi Is there a better way to do this, then? I thought this would have been the poster child for that pattern.

In case it’s useful, I should also have made clear that this is CLJS on Node, not Clojure on JVM.

http://docs.libuv.org/en/v1.x/threadpool.html might be the threadpool inside node