RouteRandomly should select the same shards within a pipeline

[sk-]

We are currently using RouteRandomly in our pipelines to query a redis cluster. However the current implementation, selects a random node for each command, which in practice for our case means that we hit all nodes in the cluster. This is not ideal as tail latencies end up being much higher, due to how the max of random variables is distributed (see https://stats.stackexchange.com/questions/18433/how-do-you-calculate-the-probability-density-function-of-the-maximum-of-a-sample).

Below it can be seen the P50, P99 and max latencies of our pipelines (they are similar in nature and size)

What do you think of adding an option or rewrite the strategy, so that it selects a random node for each shard at the beginning of the pipeline (or lazily when processing each command) then reuse those nodes for all commands hitting a specific shard.