I have an index on a and b :

If I use two consecutive $match'es in the pipeline, the index is used but only for the "a" part.
{$match:{"a":xxx}}, {$match:{"b":xxx}}

The server could optimize this case by internally merge the two match.