Where this model plausibly helps today
A few use cases map cleanly onto what decentralized identity and verifiable credentials are actually good at. Age verification is a strong fit, it's a narrow, binary fact that benefits enormously from selective disclosure, letting someone prove they're old enough without handing over a full ID. Professional licensing and credentialing is another good fit, a license or certification is naturally a claim issued by one trusted authority and checked repeatedly by many different verifiers over time, exactly the shape the issuer-holder-verifier model is built for. Proof of humanity or uniqueness, establishing that an account or action corresponds to one real, distinct person without necessarily revealing who that person is, is a plausible emerging use case as more interactions move online and need some proof that they're not automated or duplicated.
What these use cases share is that the fact being proven is narrow and well-defined, and the same fact needs to be checked repeatedly by different parties who shouldn't need to individually contact the original issuer. That's precisely the situation where cutting out the 'phone home to a central authority every time' step pays off. Where the underlying fact is fuzzy, one-off, or highly context-dependent, the benefits of this model are much less clear.