The routine for fetching all child docs for a list of parent IDs had to make multiple calls to OpenSearch to get the count of child docs for each parent. This was done to ensure we don't do a paginated search that needs more than 10k results. To get around the 10k result limit of the from pagination, I changed to the scroll pagination so this eliminated the need to make those calls to get doc counts.

The fetching of parent docs is now done using mget instead of get.

The doc reconstruct is also done in batches.