The sugar-jhbuild version of Sugar has built and updated consistently for me (on Feisty) all week, so I've gotten to spend a little more time kicking the tires and watching development in real time (as opposed to just trying to get the build scripts to work). Not that there is a high probability that I'm both wrong on details about the current state of development and that things will certainly change vastly between now and when this is deployed. But I think I've got the general idea right, and you'll probably end up better informed than you started.
The "Journal" activity (pictured above) now does enough to get a general sense of how it will work. "Journal" doesn't refer to writing. The journal is the main way students will interact with files, replacing the traditional routine of saving files in a hierarchical filesystem. It is structured like a chronological log.
The lines that say "Test" above are versions of an Abiword document. Every time I close the Write activity, it saves a new version of what I'm working on. If I want to resume working on the most recent version, I select the last one in the journal. If I want to go back to an earlier one, I can scroll up. I can also share any of these versions with my peers in various ways. I would note that, while it is almost certainly not a design goal of the development team, this could be nice for plagiarism detection, since you could see whether or not the text evolved organically or just appeared in large, perfect(ish) chunks. Also, note that turning in work in this system does not require uploading and downloading from a server. It can be shared directly over wireless.
The second-to-last item on the list is a web browsing session. When you start up your browser you can give a name to your session. It is not exactly clear how this works, because at the moment my browser activity is crashing whenever you try to navigate to a new page, but I think I get the idea. You teach kids to label what they are doing, e.g. "dinosaurs", so they can easily stop and pick up a line of research, and they can come in the next morning and share the page or sequences of pages easily with their peers.
The list of things shown in the journal can be filtered in various ways so it doesn't get too overwhelming. Also, there will be heuristics for backing up the journal to the school server and weeding out old revisions when you start running out of space. Merging different versions would seem to be an issue for future development, too. Luckily this is a pretty well understood problem for programmers.