I like this and find it interesting, but at an organizational level, it strikes me as trickier to do. A lot of the things we discover about projects and technologies have to do with its feasibility at whatever scale our real projects operate on, and don't necessarily pop-up in smaller experiments. (And sometimes they aren't really technical issues at all, but issues of "can we get most developers here to understand doing it this way.)

None of which means we shouldn't do more of this. You can learn things by trying smaller projects. It's just not a guarantee it will work in the large.