Anyway back on the sha256 ASIC issue:
Um, there is a quite reasonably serious reason why we could dump sha256 and change to something else.
Exactly the same reason for NIST's directive that U.S. government agencies had to stop uses of SHA-1 after 2010.
Once problems are found with sha256, then we'll move on to the next choice (sha512 I'd guess - or sha-3 if sha512 is considered problematic at the same time as sha256)
So anyone looking to invest 100's of thousands (or millions) of dollars into making ASIC miners is taking a much larger risk than making FPGA hardware
I completely agree with this. While I think that changing the algorithm just because there is an ASIC in the wild won't happen and would be a silly thing to do, it's completely possible that we need to change from SHA256 for another reason in the future.
The history of cryptography is full of back-and-forth of new techniques and those techniques becoming outdated and insecure for a variety of reasons. One thing that remains constant is the people who bet on their encryption technique being completely secure forever have run into big problems sooner or later.