MongoDB Index and Natural Sort Optimization - sorting

Question

What's the fastest method to sort by reverse insertion order on a capped-collection ('rf' has been sparse-indexed)
db.log.find({ rf : 'o-5556457634'}).sort({ '$natural' : -1 }).explain();
{
"cursor" : "ReverseCappedCursor",
"nscanned" : 1654468,
"nscannedObjects" : 1654468,
"n" : 4,
"millis" : 2932,
"nYields" : 5,
"nChunkSkips" : 0,
"isMultiKey" : false,
"indexOnly" : false,
"indexBounds" : {
}
}
seems like 'natural' bypass the use of the indexed ('rf') field, significantly slowing the query. Is this an expected expected behaviour? Shouldn't the 'natural' sort be computed after the find/index?
without the 'natural' sort:
db.log.find({ rf : 'o-5556457634'}).explain();
{
"cursor" : "BtreeCursor rf_1",
"nscanned" : 4,
"nscannedObjects" : 4,
"n" : 4,
"millis" : 0,
"nYields" : 0,
"nChunkSkips" : 0,
"isMultiKey" : false,
"indexOnly" : false,
"indexBounds" : {
"rf" : [
[
"o-5556457634",
"o-5556457634"
]
]
}
Hint does force the engine to use the 'rf' index but the result bypass the (reverse) 'natural' sort
db.log.find({ rf : 'o-5556457634'}).sort({ '$natural' : -1 }).hint({rf :1}).explain();
{
"cursor" : "BtreeCursor rf_1",
"nscanned" : 4,
"nscannedObjects" : 4,
"n" : 4,
"scanAndOrder" : true,
"millis" : 0,
"nYields" : 0,
"nChunkSkips" : 0,
"isMultiKey" : false,
"indexOnly" : false,
"indexBounds" : {
"rf" : [
[
"o-5556457634",
"o-5556457634"
]
]
}
}

Answer 1

Faced the same problem but found a solution.
You can create index on the fields you're mentioning in find filter with addition of the "_id":-1 field and then use sort({"_id":-1}).
Helped me.

Answer 2

It looks like the query optimizer is doing the wrong thing when you add the sort.
Can you try adding .hint({rf :1}) to the query to see what happens?