However, keep in mind that all these messages need to be correlated and subject to business logic and validations. For example, the business logic might dictate that one cannot record a temperature reading to the blockchain if the chain-of-custody does not show you as the current possessor of the item.
Alternatively, the business logic may say that you cannot blend a non-blendable lot.
There are thousands of business rules that need to be enforced along the way.
With the (say) PGP approach, you would need to ADDITIONALLY trust that the entity applied these business rules correctly. Also in this approach, there is no single version of the truth. So, every entity maintains its own version of the truth leading to huge synchronization problems in addition to trust problems.
My point is that this is just not achievable using traditional DBMS: the original premises ("trust model" and incentive model) of those systems are very different.
Please correct me if I am wrong somewhere.
I don't really understand that. The "who" can be authenticated using digital signatures (say, a PGP signature for instance, or a certificate signed by the authority in change). This is exactly the way a blockchain would work, making a transaction means signing it with your private key. As long as your private key remains private nobody can impersonate you (but that's not revolutionary blockchain technology, that's goold old seventies asymmetric crypto technology).
As for the "what" I'm not sure what you mean by that. If somebody (the "who") digitally signs a message or transaction and its contents (the "what") gets modified somewhere along the way then the signature will fail to validate. On the other hand if I get a signed message by entity A saying something that turns out to be false then it's proof that A lied, was compromised or made a mistake.