I'm thinking of taking advantage of the sliced scroll api. Based on what I've read, I'm not too sure if using the sliced scroll api will have benefits given that I route all documents.
From the scroll slicing section, it says that it will divide the work to be done first among shards and then within an individual shard.
"By default the splitting is done on the shards first and then locally on each shard using the _uid field with the following formula: slice(doc) = floorMod(hashCode(doc._uid), max)"
So lets say, I route my documents during indexing then it should go to 1 shard. If I then attempt to scroll with a routing parameter then ES should scroll against that shard. If I then attempt to slice the scroll with routing will it know to look only on one shard and just proceed to divide the work within that shard? Or will it ignore the routing value and attempt to scroll against all the shards for index.