It is straightfoward to build systems which derive their state from the audit trail instead of building the audit trail in parallel. That is what event sourcing is.
I was attempting to emphasize the absurdity of any software system being “absolutely correct at all times”. I don’t believe such a system can exist, at least not in such strong terms.
What's important is that the audit trail can be replayed to derive the state of the system - and preferably in such a way that investigators can determine what _would_ have been seen by someone using it on a specific day at a specific time. Whether the system is free from bugs is a different matter - no system is, which is why deriving state from the audit trail instead of a parallel process which is guaranteed to diverge is so important!
