please help me with my flawed logic on how to observeChanges in a clustered multi-server environment (with oplog-scaling).

cluster_observeChanges.js

// 200k entries and more
HugeCollection = new Meteor.Collection("HugeCollection");

// convenience collection to hold the computed stats of HugeCollection
StatsForHugeCollection = new Meteor.Collection("StatsForHugeCollection");


/**
 * Observer
 */
if (Meteor.isServer) {
    var HugeCollectionObserverHandle = HugeCollection.find({}).observeChanges({

        /*
         this works perfectly on a single instance server but
         when scaling out horizontally with eg. 4 cores, each stat inc.'ed 4 times
         */
        added: function(_id, fields) {
            // example stats, real scenario is much more complicated
            StatsForHugeCollection.upsert('total_sum', {$inc: {total: fields.value}})
            StatsForHugeCollection.upsert('total_count', {$inc: {count: 1}})
        }
    });
}


/**
 * Pub / Sub
 */
if (Meteor.isServer) {

    // the stats are really important for the clients,
    // the actual HugeCollection data is not
    Meteor.publish('publishStats', function() {
        return StatsForHugeCollection.find()
    });
}
if (Meteor.isClient) {
    Deps.autorun(function(c) {
        Meteor.subscribe('publishStats');
    });
}