in short if the hash function is to be replaced it must be replaced with something different. for example one of those sponge based constructions instead of MerkleDamgård, like Keccak (standardized as SHA-3).
And this I guess is why when we're seeing proposals, they're for different/new hash functions? Although until I saw this I didn't know about this SHA-3 family of standards. I don't trawl all the proposals but that's what I recall (or don't recall) but I always thought that when, if at all, Bitcoin decides for a major change, it will only be to hash funtion, not hash algorithm.
Are you aware of anything suggesting otherwise?