There's more to the world than science. I donate spare CPU cycles to the pypy buildbot.
While I think OGR-28 is neat, what advantage is there in taking less than 15 years to knowing we have a perfect ruler? After all, we have a ruler now, we just don't know if it's perfect.
Regarding Folding@Home, papers like http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3673555/ suggest that dedicated GPU hardware is the way to go these days. Reviewing Pande's papers at http://folding.stanford.edu/home/papers suggest that the limiting factor to understanding folding is analysis capabilities, not data generation.
Cosmology@Home does not seem to be that active. Quoting from a 2011 message at http://www.cosmologyathome.org/forum_thread.php?id=590 " Anybody who crunches here is probably just wasting his machines and electricity for absolutely nothing.", and the welcome letter at http://www.cosmologyathome.org/wandelt_letter.php says "As a further incentive for people to participate we are considering offering the Cosmology@Home Prize for the owner of the computer that calculated the model that best fits the data as of the 31st of December 2009."
Given that the Amazon EC2 spot price for a computer more powerful than my desktop is about $0.01, it's hard to be enthusiastic about spending my time. See for example the comment at http://en.wikipedia.org/wiki/SETI@home :
> it has not been demonstrated that the order of magnitude excess in computers used, many outside the home (the original intent was to use 50,000-100,000 "home" computers),[11] has benefited the project scientifically.
While that comment is unsourced, I see no reason to disagree with it.