Currently, the most proof-of-work chain for Bitcoin is the one that most SHA256 work. If a hard fork is done that allows for continuation with some other algo, then the most proof-of-work chain would be defined differently. Maybe the definition would consider SHA256, but not exculsively. Even if the attacker gets more proof of work for SHA256, they would have a hard time getting more than Bitcoin for some other algo I assume. It depends on how it's done, so not sure, but either way it would be an ugly situation, and something we want to avoid.