From the mailing list:
db.test.insert(

{ parts : ["part1", "part2", "part3", "part4"], priority : 2 }

);
db.test.insert(

{ parts : ["part4", "part5", "part6"], priority : 5 }

);
db.test.ensureIndex(

{ parts : 1, priority : -1 }

)

// Uses an index
db.test.find({$or : [{ parts :

{ $all : ["part1", "part2"] }

},

{ parts : "part4" }

]}).explain()

// Does not use an index
db.test.find({$or : [{ parts :

{ $all : ["part1", "part2"] }

},

{ parts : "part4" }

]}).sort(

{ priority : -1 }

).explain()

// Uses the index, but doesn't produce expected explain output
db.test.find({$or : [{ parts :

{ $all : ["part1", "part2"] }

},

{ parts : "part4" }

]}).sort(

{ priority : -1 }

).hint(

{parts: 1, priority: -1}

).explain()

// Here's the output (it's NOT in the form of the typical $or-like explain plan)
{
"cursor" : "BtreeCursor parts_1_priority_-1",
"nscanned" : 7,
"nscannedObjects" : 7,
"n" : 2,
"scanAndOrder" : true,
"millis" : 0,
"indexBounds" : {
"parts" : [
[

{ "$minElement" : 1 }

,

{ "$maxElement" : 1 }

]
],
"priority" : [
[

{ "$maxElement" : 1 }

,

{ "$minElement" : 1 }

]
]
}
}