My personal philosophy with Node.js is to use it for small, standalone network servers that complement the rest of my stack - basically anything that needs to handle large amounts of I/O. Everything I've written with Node.js so far has been just a few hundred lines of code.
So for templating, database interaction and so on I'll keep using Django. I'll use Node for stuff like comet/WebSocket pubsub services, handling file uploads, rate limiting API proxies, webhook dispatching, interacting with slow or streaming external web APIs, etc.
GitHub are using Node in this way at the moment: https://github.com/blog/678-meet-nodeload-the-new-download-s...
Many of my problems with node are simply a matter of taste and I'm not trying to discourage the use of Node.js; its pretty good for quick and dirty web services but, in my personal experience it has not been worth the trouble for large projects.
Quora thread for which companies are using node: http://www.quora.com/Node-js/What-companies-are-using-Node-j...
Some things are a bit less mature than one might desire, but all in all it's a pretty brilliant piece of technology and I find myself using it more and more.
I wouldn't use it as my main webserver, but it's really nice for this part of it.
edit: We have a thick client using Cappuccino and we talk to our Node web service using JSON. The only templating on the server is for a few HTML emails, password reset page, and the email verification landing page.
I was disappointed to learn javascript does not support tail recursion. Kind of makes it a 2nd class functional language.