twomodelsForCollection.js

var mongoose = require('./../mongoose');
mongoose.connect('localhost', 'testing_twomodelsForCollection');


var schema1 = new mongoose.Schema({
    name: String
}, {collection: 'mystuff'});

var schema2 = new mongoose.Schema({
    age: Number
}, {collection: 'mystuff'});

var A = mongoose.model('A', schema1);
var B = mongoose.model('B', schema2);

mongoose.connection.on('open', function () {

  console.log('creating..');
  A.create({ name: 'first' }, function (err, a) {
    if (err) return console.error(err.stack||err);

    B.findById(a, function (err, b) {
      if (err) console.error(err.stack||err);
      console.error('found an A from B!', b);

      b.age = 45;
      b.save(function (err) {
        if (err) console.error(err.stack||err);
        console.error('saved!');

        A.findById(b, function (err, a) {
          if (err) console.error(err.stack||err);
          console.error('finished', a);

          mongoose.connection.db.dropDatabase(function () {
            mongoose.connection.close();
          });
        });
      });

    });
  })
});

I figured out, how I could do something in that direction, but I'm not sure how efficient it is. I think I'm now letting mongoose take an additional trip to the mongodb to populate the fileId field, regardless if its set/existent/null;
A.
findById(id).
populate("fileId").
run( function(err, a){
if (a.type==="b"){
var b=new B;
b.init(a._doc);
}
//a would'nt know about the field fileId (and not have it populated)
//b would know about the field fileId and have it populated
})