The breakthrough claimed is that only one side (the router side) is equipped with a qubit detector, and end nodes only have quantum transmitters, and therefore the cost of deployment for each end node is much lower. This tech doesn't change the basic limitations of what quantum cryptography requiring trusted nodes can achieve: It only secures fiber links, and every router along a path still has to be trusted (ISPs have to be trusted and incorruptible, for instance). It also doesn't help prevent typical hacking, where nodes are attacked, not network links.

See http://www.youtube.com/watch?v=UVzRbU6y7Ks for the quantum key exchange protocol. The protocol doesn't need a quantum back-channel. What has been "declassified" about this approach?

Whenever you purchase something online and you hit the ‘buy’ button and your computer seems to buffer, taking its sweet time to present you with the order confirmation page and you begin to doubt that the order went through successfully, “that’s because of the cryptography,” says Hughes. It takes time to create a secure line to transmit sensitive information, like your card number, between your laptop, eBay, and your bank. But “in our case that just wouldn’t happen,” says Hughes, “in principle [our invention] could speed up the Internet.”

Bad journalism. That quote is probably taken way out of context. Website delays usually have very little to do with cryptography, and everything to do with the web apps and backend databases or credit card processors being slow to process orders. The slow part of SSL encryption (key exchange) isn't that slow, and is probably already done before you submit an order, if the site uses a ssl session cache.