leverage in-flight requests of other replicas to possibly overcome shard query errors

[Dieterbe]

fix #1817

one interesting case is what happens when an error comes in, and there are more replicas available for querying, as well as a request in flight (you can have this with replicas >= 3)

we could do two things here:

1. wait until the in-flight response comes in, and only when an error comes back for it, try the 3rd replica
2. initiate request to 3rd replica, resulting in 2 in flight requests (and a total of 3 for the given shard)

I've opted for number 2.

[shanson7]

Minor comment. Overall looks much cleaner!