Our documentation on multikey behavior and bounds creation is incomplete. Also our bounds w/the new planner are probably not as tight as they could be.
Potentially unexpected scans with compound indexes when some fields are multikey
When choosing multikey index bounds, never choose a superset if a subset is available
Doc index bounds behavior for multikey indices
Query on compound multi-key indexes lost performance between 2.4.6 and 2.5.5-pre-