<sarcasm>
I think you meant <pissy> :)
A dedicated server to receive crash reports automatically would probably
take a lot of abuse and you'd have to sift through the real reports vs
the sheer volume of people who just have bad hardware.
You can't mine the data until you have it, and having it is not an
obligation to mine it. It's just an opportunity.
There's probably an unbounded amount of work that could go into mining
it, but I'd suspect that you could get a pretty good ROI just by going
for easy signals-- a rash of crashes on chipset foo on kernel x.y.z, for
example. It sounds to me like a pretty decent idea, modulo privacy
concerns.
-Ed