For the same reason they didn't use SHA1024 or SHA2048 or SHA4048 or SHA1000000000000000000000000000000000
No. SHA512 and Whirlpool exist, are well defined, well supported, well analyzed, and they exist for a reason.
There are lots of theoretical attacks that can be done against it, but if a program or new math proof can half the amount of time it takes to crack it,
Reversible computing techniques 'cheat' around the entropy limit. This means they can reach effective speeds far, far beyond what are possible with current computers, as they are effectively capable of performing nondeterministic operations.
You are basically betting the entire economy (if you believe bitcoins will succeed anyway) on no one developing a means to halve the effective bit length as has been done with e.g. AES.
It's careless.
are we really worried about the encryption taking 100 billion years to crack but now with this new attack (insert math,attack,flaw) it's only going to take only 1 billion years to crack? How about a million years? Even one-hundred thousand years?
Ten years, assuming only minor flaws in SHA256.
If there is a major flaw (again, see the push for SHA-3) there is a much more serious problem. There does not appear to be a clear mechanism for handling a compromise of the basic algorithm, and there should be.