Where does the trust boundary sit when modeling real-world events on-chain?

I’ve been thinking about how crypto systems could handle irreversible real-world events like death or permanent key loss.

Most designs assume the signer is always available, so when that breaks, assets just become inaccessible. I came across an open-source project called Afterchain that tries to model this using an on-chain execution layer combined with external attestations (proof-of-death, multisig oracle, etc).

Repo: https://github.com/Afterchain/afterchain-protocol-public

The execution logic itself seems straightforward, but the attestation layer feels like the hard part. Even with things like ZK proofs, you’re still relying on some form of trust in how real-world data gets introduced.

Curious how others think about this, is this basically an oracle problem, or is there a cleaner way to handle real-world state changes in a trust-minimized system?