[simianwords]

we had something like this to scale out for higher throughput. just in the 10's of thousands requests per second required 100+ nodes simply because each query would have a expensive scatter and gather