The internal representation for the sort order of a plan cache key as defined in canonical_query.cpp:269-276 is ambiguous because it does not use any separators for multiple keys.

These two queries produce the same cache key and could therefore incorrectly use a wrong cached plan:

db.coll.find( {a: {$gt: 1}, b:1} ).sort( {a:1, b:1} )
db.coll.find( {a: {$gt: 1}, b:1} ).sort( {aab:1} )

Both produce the internal key: an[eqb,gta]~aaab.

This could be avoided with a separator between multiple sort keys, e.g.

an[eqb,gta]~aaab
an[eqb,gta]~aa,ab