For $unset, the RHS value is not only ignored, but also not stored in the update tree. The question is should $unset accept all kinds of values, or should "1" be the only valid input and should be validated during parsing. Beginning to validate in a new version may be tricky, but perhaps we should track the query shape/stats differently for all the various inputs so we can get a sense of how often the non-1 case is "in the wild"?