Yeah, that the idea. The trick is the ordering of npubs so that (n+1) npub can announce that (n) npub is no longer valid. Similarly, someone can determine which events were valid and at which point they are invalid. Likely will require an open timestamp.
Discussion
wouldn't there be a case where you could announce validity of npub (n) with (n-1) or any prior npub that is still good?
TBH, I haven’t thought it through completely. The question, I believe is specifying the validity period or state of a prior npub with something like “expired”. The problem is that the prior npub, if compromised, could specify a bogus n+1 pub. I think this is an open timestamp-like problem.