The projects can be of very varied nature. First rule: don't fear redundancy! You shouldn't strive for a software revolution in 2 hours. And here are a few more tips
This is what most projects will probably look like. "Develop a player > 2 pong game", "Develop a usb-stick portable TODO application in pure JavaScript? with jQuery".
On this type of project, you set yourself the goal to answer questions about the inner workings of some software, if it has poor documentation for the angle you're interested in. This can be to satisfy your own curiosity, to support some hackish functionality you want to add, etc. The "deliverable" here could be the very documentation you produce along the way (which you can post to your blog, for example). This could also be useful to others who'd then want to themselves pursue further (see also "adding feature to open source program" bellow).
This can of course be done in teams: each tackles a specific dimension of the program, and post their findings as they go along on a wiki page, for example.
We all have a list of features we'd like to see come to reality in our favorite software. This can be a good opportunity to finally implement some of them!
Clusterify can also offer the opportunity, for open source software authors, to interest people in their project: post a clearly defined piece of job to be done, and help people who join in realizing the task.
If you're out of ideas, you can always join projects created by others, you know :P Yet, if you want some "random seed" to get the idea-generation process started, you can take a look at problem solving sites. Here's a great list.
