Thanks, that's a really good suggestion. I hadn't considered a RoaringBatchIterator. I'll try this approach and benchmark it against the current forEachInRange implementation to see how they compare in practice.
I'll let you know what the performance difference looks like.
