we pick the server with the least queries 'in the air' (and within those,
the one with the lowest `order`, and within those, the one with the lowest latency).
-Another policy, `firstAvailable`, picks the server with the lowest `order` that has not
-exceeded its QPS limit. For now this is the only policy using the QPS limit.
+Another policy, `firstAvailable`, picks the first server that has not
+exceeded its QPS limit. If all servers are above their QPS limit, a
+server is selected based on the `leastOutstanding` policy. For now this
+is the only policy using the QPS limit.
A further policy, `wrandom` assigns queries randomly, but based on the
`weight` parameter passed to `newServer`. `whashed` is a similar weighted policy,