The auto-updating template stuff is cute, but unless your goal is apps that max out around the complexity of a todo list, there's a ton of performance pitfalls (c.f. Ember.js), especially in mobile browsers. A lot of apps are that simple though, so I don't want to knock Meteor (or Ember) here -- that's a useful niche to be in and anything that makes development faster/easier is great!

A Blossom app would simply use the server capabilities and data updating stuff in Meteor, but in the context of a normal, high-performance Blossom-based mobile UI with statecharts, hardware-accelerated surfaces and animation, and yes, GPU-accelerated Canvas rendering (although Blossom now supports old-style HTML/CSS views as well).

Mobile apps written with Blossom have virtually the same amount of resource control as a native iOS app, and to my knowledge, is the only mobile app framework with 100% JavaScript controls that can say that. (Appcelerator's Titanium is the other way to go – wrap the native controls with a JavaScript API.) Not to mention the ability to run on Android and iOS browsers from the same code base, or deploy via PhoneGap, or target desktop browsers or tablets with effectively native performance.

Meteor is really neat, I did something similar to at least the data syncing part two years with a tech I developed called HubSync, which kept a SproutCore datastore in sync across clients. The net effect was the same -- no server code, data stays in sync automatically, and you only ever dealt with the client API. (It could do a few other things, like still function offline, and was also shared nothing on the server side with Cassandra or BigTable -- I've seen serious problems with Mongo scaling in the past (yay, global lock!), and I have no idea what the offline story is for Meteor.)

But still, very cool tech overall, and the team looks like they've got some great ideas. Bravo!