Hi all, I see that DocAndScoreQuery's Weight in AbstractKnnVectorQuery is cacheable now. Is this correct?
I doubt because if we use KnnQuery as a part of a filter (for example and(someFilter, or(knnQuery1, knnQuery2))), we get DocAndScoreQuery cached. But as far I understand after index reopened, we may not be able to use this cache because we have to use exactly the same reader for creating weight: https://github.com/apache/lucene/blob/cc3b412183c7acbb4a6c498e9babffba4e7d1f28/lucene/core/src/java/org/apache/lucene/search/AbstractKnnVectorQuery.java#L369