I'd be interested to see the man-hours-needed : people-helped ratio when it comes to user accessibility. There's certainly a threshold where it simply doesn't make sense to put in a man year's worth of work to cater to a proportionally very small percentage of the population.