On the server side, there is one working thread per shard.
As you said, scroll ids can change so if you want to parallelize logic on client-side, you still need to make sure that the call to get the next page is not parallelized.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.