"Truth" of course is a red herring. "Claim" is better. And, "digitally signed claim" is better yet.
Then to answer your Q: a good sharding approach should let you see all digitally signed claims (including when those claims were made) with probability --> 1.0. I'm framing this probabilistically because many sharding approaches rely on that definition. (And even non-sharded blockchains like Bitcoin itself.)
Also, I had the impression that a blockchain is a specific type of data structure.
So, what you are describing seems like a way to accomplish behavior of this kind of data structure with a different one so you may get bettee scaling out of it.
As you'll see in my other comments (and articles on it), "blockchain" is better described as a field with a set of related goals for technology artifacts, rather than a specific data structure [1]. I frame it as: it has blockchain characteristics if it's decentralized, immutable, and assets [2].
This is a much healthier framing, because it doesn't constrain the goals to a particular approach (e.g. a particular data structure).
Then to answer your Q: a good sharding approach should let you see all digitally signed claims (including when those claims were made) with probability --> 1.0. I'm framing this probabilistically because many sharding approaches rely on that definition. (And even non-sharded blockchains like Bitcoin itself.)