Surely, quite a lot of time a design can be tweaked in its current version to make some great changes which drastically improve upon the usability. But on the contrary I think that Windows has reached a stage where some minor tweaks in its current incarnation won't make a lot of difference. It can only be optimized not radically improved until it is completely though from bottom up.
You might like to read about the concept of "Local Maximum"[1] on 52 weeks of UX. Here's a excerpt from the post.
Do you ever feel that your design has become stale and that despite your making lots of little changes to it over time without any big overhaul there is just no way to drastically improve it? If so you’ve probably hit what Andrew Chen calls the “Local Maximum”.
While i'd love to see this and it is undeniably beautiful, its not really what i'd call a practical solution. This makes me wonder, is it only beautiful because its a concept? It only looks so good because its free from practical restriction, what would it look like if it was made into a real world application?
This also makes me ask, is the only way to radically change UI to come up with a completely new from scratch paradigm, and if so, is this why we have metro?