Umm .. why does this list only contain clones of commercial softwares? I would rather say that high priority projects are the linux kernel, git, myriad python and ruby interpreters etc which are widely used and do contain some design innovations.
How about something completely original? How about we start racing the commercial sector to a worthwhile goal like computer vision, or decent speech recognition. Anything more inspiring than a "Skype replacement" will do fine.
We just need a few more things where the open source version is the dominate player, instead of a clunky, half working "free replacement" for commercial software that wasn't very good to begin with.
While Octave is good numpy/scipy/matplotlib are also free (in both meanings of the word) and great software. There is also Sage.
So get started writing machine learning libraries in Python!