If hypothetically bitcoin fell to <$5 and stayed CONTINUALLY under $5 for 30 days (long enough for miners to get next electric bill) you likely would see some behavior change.
That's the key right there. Most people will stick it out at least until the next power bill hits and maybe even for two bills, hoping to see a turnaround in price or drop in difficulty that will keep mining profitable. There's definitely some people out there that will continue to mine regardless, but I bet some of the power goes dark after a while. For example, I'm running 11 GH or so and it's costing about $500/month with a great .085 cents/kwh power rate. If it gets unprofitable for me I'll still probably keep a GH or two running to help the network, but I can't afford to throw $400-$500 at it each month with no return for very long.
You mean you're not one of those Folding @ Home guys, whose idea of fun is to spend their extra money on hardware to crunch numbers?
I'm not such a (what's the most derogatory word for geek or nerd?) that I'd spend even $20 on electricity for some silly distributed computing project.
I'd rather take my wife out to dinner or ANYTHING I can actually enjoy. I guess I have a life...
More like donating to science at best. If you really want to donate to charity, find someone LOCAL who needs the money and help them out. Charity from-a-distance is like government from-a-distance: fraught with waste. Charity is best done at home.
Anyhow, the chance of "donating $20 to science via distributed computing" doing any actual good is pretty slim. You'd be better off donating the $20 to some cause you believe in -- a scientific foundation if you prefer.
There's a lot of overhead in distributed computing projects, and only certain problems can be broken down into segments where reliability and resiliency aren't important. Remember, any piece of the puzzle can go "poof" if the person with that slice decides to quit. So the pieces of the puzzle can't depend on each other too much.
Many number-crunching scenarios could never be written for a distributed computing platform. Too much memory is required, the calculation can't be easily split up, etc.