If "don't make mistakes" was a viable solution, we just wouldn't have a whole family of problems.

> you probably should include a timestamp, an expiration for the key, a hash of the previous message so people can tell what is supposed to supercede what, etc.

An expiration for your current key doesn't _really_ help the problem of "I don't know where to get the next key from". A blockchain _in theory_ formalises the "disseminate your public key through this accessible place" (which you've defined as twitter, I might define as a CA registry that comes with browser and someone else might define as a blockchain), but this solution doesn't actually solve the problem of "the public key is different, which one do I use?" which _is_ what a blockchain does, except in reality, it's useless because when these things diverge you almost certainly look elsewhere for trust.