Switching Bitcoin protocol will be extremely difficult and risky. If petaflop power of currently implemented ASIC hardware is lost, the network would become so weak that it'd be vulnerable to anyone who can afford to spend $50ish-million on developing ASICs that use new protocol.
So switching is viable, but dangerous for the network until overall computational power is restored to the point where no single entity can easily amass 51% of hashrate. The 'problem' will worsen as time goes on and more SHA256 ASIC are added to network.
Luckily, the market has already created a solution to this. It's called Litecoin (only big non-SHA-256 coin at present). As for ASIC miners, their expected ROI point is usually 3 - 6 months out at most anyway; and most are aware SHA-256 will not live forever. The market will hedge its bets with a big shift to LTC if necessary, but really I doubt the actual cracking of it will be a big issue. It's unlikely to occur and be used in an attack. More likely, it will be published (maybe even with a couple weeks of heads-up lead time to Bitcoin devs, since that's the single biggest implication of it) with the details held secret until the hash function is updated and mining strength is strong (which would happen quickly, as even CPU mining would again be viable, and the circle of life of a crypto's hash function continues).
Also, please explain your logic in thinking a 51% attack is any more viable while a new hash function is in it's infancy, and everyone goes back to CPU, then GPU, etc. It seems to me that this is perhaps actually desirable, for the decentralization and prevention of 51% accumulations. I believe Satoshi would agree with me.