Also see http://lwn.net/Articles/375550/ from the comments and its associated thread.

I wonder if this article will motivate hackers to change/complete some of the particularly old untouched or unfinished code mentioned.

Such a technique might be a good way to find older code which could be improved upon or upgraded.

I wonder if most projects would have a similar curve, and if parts of the curve had different properties (bug count, difficulty to change, etc.)

I think projects of that age (18 years?) and size (~10M LOC?) which are still so much in flux are rare. Of course, Linux has a gigantic number of people working on it, and a large part of the codebase is hardware drivers - and a lot of new hardware is released all the time. In that light, the graphs aren't overly surprising, but I can't think of a comparable project.