You don't need a full 1:1 mirror of the strands, you can probably filter off a lot of the traffic that you don't need to spy on (e.g. TCP overhead) compress it and use a fraction of the bandwidth.
The worse thing is that when one of the developers tried to write a non-GPL version of BusyBox they were criticised for "taking away the only relevant tool for copyright enforcement of the Software Freedom Conservancy group" - they're as bad as patent trolls.
+1 If you want your code to be used you need to use a less restrictive license and trust organisations and their developers to do the right thing. Open sourcing everything is not always the best solution and if you force them to open source their changes the chances are you're going to get some internal branch open sourced hidden on some corporate site somewhere, that's not going to give anything back to the community because somebody is going to have to pull those changes back and support them. If you want to build quality OSS then you need to engage the development community, not put a gun to their head and force them to open source.
Lots of companies open source their products under the AGPL with a commercial license they sell. This allows them to say their products are 'open source' and look good, but in reality it hampers their own product. If I wanted to commercialise that product and make changes, I can open source the changes, but I won't sign the original companies CLA which means they can't take my code and use it in their commercial version.
This lawsuit will be interesting though, as it will probably set a precedent on how you can use GPL software in commercial software.
Our uni has been working with a new startup based in the UK working on a completely new LMS kind of system. It's really cool and well designed and takes on features from popular internet services and builds them into education (ala. Facebook, it's also really well designed, I think there are some ex-Apple/Google people working for them). The company though has also been working with Students / Lecturers and Administrators so they've done a good job of integrating with our systems (disclosure: I'm the LMS Technical Manager). But it's also built from a learner and teacher perspective and they guys we've given it do really like it.
If anyone wants to talk about it you can email me at rusdyas {at} gmail.com (kind of non-disclosure, over at least the details anyway)
I've actually been using Ember for quite a while now and it's formed the basis of one of our products.
Once you get over the massive learning curve with Ember and keep within the 'emberisms' (i.e. the way Ember does stuff), it's an incredibly powerful framework. In terms of hiring people to code in Ember, it's much easier to onboard people once you have a few developers that really know ember through and through. The biggest problem i've found is that for a long time the API's changed around a bit (they are fixed now) and that meant that sometimes if you're on StackOverflow, the fix may not work anymore.
Having used Backbone, Angular and Ember; I would describe Backbone as the past, Angular as the present and Ember as the future.
Ember's got some pretty big people using it, Zendesk's new agent interface is built in it, Discourse, Square and (indirectly) Apple (iCloud uses Sproutcore which is where Ember originated from).