Collection scans are sometimes faster than index covered queries. When accessing data from an index, the scan creates a BSON object for each document. We could alleviate the cost of this operation for duplicate documents by caching the object.