This problem might be hard to duplicate, but basically I have a unique compound index that may occasionally be too large. Unfortunately I'm not seeing any notification of this problem when the index is created, and only know it's a problem because a.) uniqueness is being violated, and b.) queries with the index are not returning all the records I expect.

My class structure is basically like this:

class Simulator
  has_many :profiles
end

class Profile
  field :parameter_hash
  field :proto_string
  belongs_to :simulator
  index ([[:simulator_id,  Mongo::DESCENDING], [:parameter_hash, Mongo::DESCENDING], [:proto_string, Mongo::DESCENDING]]), :unique => true
end

Now for the weirdness. I create a Profile with a very large proto_string. Let's call this record 'profile'. This record can be found with

Profile.where(:simulator_id => profile.simulator_id).count 
# => 1
Profile.where(:parameter_hash =>  profile.parameter_hash).count
# => 1
Profile.where(:proto_string =>  profile.proto_string).count
# => 1
Profile.where(:parameter_hash =>  profile.parameter_hash, :proto_string =>  profile.proto_string).count
# => 1

but not with:

Profile.where(:simulator_id => profile.simulator_id, :proto_string => profile.proto_string).count
# => 0
Profile.where(:simulator_id => profile.simulator_id, :parameter_hash => profile.parameter_hash).count
# => 0
Profile.where(:simulator_id => profile.simulator_id, :parameter_hash => profile.parameter_hash,  :proto_string => profile.proto_string).count
# => 0

My understanding is that this behavior arises because simulator_id is the first piece of my index, so basically all the queries that succeed are just not using the index. So this leads me to a couple of questions:

During development/testing, how could I foresee/test for this problem? In production, how can adjust for this? I believe mongodb has a strategy for dealing with this by appending '.hint({$natural: 1})' to the end of a query where this is suspected. Is there anything similar I could do with mongoid?