Description
The BlockJoinScorer temporarily saves the child docids and scores in two arrays (pendingChildDocs/pendingChildScores) for the current block (parent/child docs) being processed. This is only need for ToParentBlockJoinCollector and in the case that this collector isn't used then these two arrays shouldn't be used as well.
I've seen cases where only the ToParentBlockJoinQuery is used and there are many child docs (100k and up), in that case these two arrays are a waste of resources.