if SHA-256 is partially broken, and its proposed to swap one of the 2 algorithms with a newer replacement such as say Keccak (SHA-3), you'd see some resistance from large ASIC investors as it's no longer feasible to use their infrastructure once the swap happens. We'd be back to GPUs or very likely the swap would be announced for a future date (a few months), giving enough time for new ASICs to be developed that run the new algorithms.

The urgency of the swap would largely depend on how much SHA-256 is broken, if difficulty can't be set hard enough or becomes too closely to being able to break the address space (https://bitcointalk.org/index.php?topic=107172.0), then urgency would be high.

Either way an interesting and potentially volatile situation.