AFAIK the reason web apps have become so popular is because they load quickly and don't require the user to manage installation and updating. IPFS would achieve the speed through caching and the installing/updating process with its namespaces feature.
Immutability: with HTTP you have to save things you care about because they might disappear/move in the future. With IPFS you could pretty much always rely on at least Google holding a copy of something in the long term.
Basically the idea is to remove the concept of local vs remote files at the OS level. So as a user of the OS the file browser and the web browser are the same application. They shouldn't even be able to tell if something is on their computer or not. This could be implemented on HTTP but only as a prototype.
Imagine a social media, where if you're in a office and the connection to the general internet dies, all communicating within the office still works.