I guess I'm having problems seeing where this algorithm is necessarily all that difficult and can't be sped up in some manner where a dedicated coin generator wouldn't have a huge advantage over a normal ordinary client in terms of the creation of Bitcoins.
Sure, you could build specialized hardware that computed SHA256 hashes faster than ANYBODY!
And then you'd be able to generate most of the bitcoins.
So: how much would it cost you, in dollars or euros-- and don't forget to pay yourself for the time needed to design and build such a system?
Because you'd be able to generate, at most, 50 bitcoins every 10 minutes or so. Soon after you started generating, all the other nodes on the network notice that bitcoins are being generated quicker than normal and increase the difficulty of the problem being solved (you've gotta twiddle the block header more times to find SHA256Hash(block_header) < target_value ).
If Bitcoins become even more wildly popular, then maybe people will think they can profit by building or renting specialized hardware. That would be a very good problem for Bitcoin to have.