MongoDB query executes in 1ms on mongo-shell but takes 400ms and more on NodeJS - node.js

Question

I have a large MongoDB collection, containing more than 2GB of raw data and I use a very simple query to fetch a specific document from the collection by its Id. Document sizes currently range from 10KB to 4MB, and the Id field is defined as an index.
This is the query I'm using (with the mongojs module):
db.collection('categories').find({ id: category_id },
function(err, docs) {
callback(err, docs.length ? docs[0] : false);
}).limit(1);
When I execute this query using MongoDB shell or a GUI such as Robomongo it takes approximately 1ms to fetch the document, no matter what its physical size, but when I execute the exact same query on NodeJS the response time ranges from 2ms to 2s and more depending on the amount of data. I only measure the time it takes to receive a response and even in cases where NodeJS waits for more than 500ms the MongoDB profiler (.explain()) shows it took only a single millisecond to execute the query.
Now, I'm probably doing something wrong but I can't figure out what it is. I'm rather new to NodeJS but I had experience with MongoDB and PHP in the past and I never encountered such performance issues, so I tend to think I'm probably abusing NodeJS in some way.
I also tried profiling using SpyJS on WebStorm, I saw there are a lot of bson.deserialize calls which sums up quickly into a large stack, but I couldn't investigate farther because SpyJS always crashes at this point. Probably related but I still have no idea how to deal with it.
Please advise, any leads will be appreciated.
Edit:
This is the result of db.categories.getIndexes():
[
{
"v" : 1,
"key" : {
"_id" : 1
},
"name" : "_id_",
"ns" : "my_db.categories"
},
{
"v" : 1,
"key" : {
"id" : 1
},
"name" : "id_1",
"ns" : "my_db.categories"
}
]
I also tried using findOne which made no difference:
db.collection('categories').findOne({ id: category_id },
function(err, doc) {
callback(err, doc || false);
});

Answer 1

My guess is the .limit(1) is ignored because the callback is provided early. Once find sees a callback it's going to execute the query, and only after the query has been sent to mongo will the .limit modifier try to adjust the query but it's too late. Recode as such and see if that solves it:
db.collection('categories').find({ id: category_id }).limit(1).exec(
function(err, docs) {
callback(err, docs.length ? docs[0] : false);
});

Answer 2

Most likely you'll need to have a combination of normalized and denormalized data in your object. Sending 4MB across the wire at a time seems pretty heavy, and likely will cause problems for any browser that's going to be doing the parsing of the data.
Most likely you should store the top 100 products, the first page of products, or some smaller subset that makes sense for your application in the category. This may be the top alphabetically, most popular, newest, or some other app-specific metric you determine.
When you go about editing a category, you'll use the $push/$slice method to ensure you avoid unbounded array growth.
Then when you actually page through the results you'll do a separate query to the individual products table by category. (Index that.)
I've written about this before here:
https://stackoverflow.com/a/27286612/68567