A time or engagement based restriction would prevent this, i.e. having 3-5 minutes to edit the tweet, at which point the edit button is locked. Revision history would still show. "Undo Send" a la Gmail, but for tweets.
Except in a distributed system like Twitter (including client and server) there is no single timeline, and amateur digital forensics will erroneously say "aha, but you retweeted it before it was edited"
Perhaps, but you are talking about creating rather complex machinery in order to support a tiny feature. If the only argument in favor is engagement statistics (would those take edits into consideration as well?), I certainly see why Twitter doesn't care too much.