I mean it's not an insurmountable problem though. As long as likes and RTs are tied to a specific version of a Tweet and you can see the edits and associated likes it's fine.
Even setting aside the UX issues in this, I think this is underestimating the complexity. I don't know anything about Twitters infrastructure but obviously we are not talking about a single postgres instance here. Effectively turning every tweet into a linked list with connected retweets, likes etc. is a significant data model change for a system of this scale.
That just makes it even more confusing. So you edit your tweet, but all the people who retweeted are still showing the wrong version and you can't do anything about it.
It kind of makes the situation worse than it is now.
It's not the "wrong" version. It's the original version, as per when they quoted it. There's really no other way it could work.
When you quote someone, you repeat what they said, not what they might repeat some time in the future. When you quote a book, you write down what's written in the book. Not what the author writes in the next edition.
You may in time edit/remove/amend your tweet to comment on further changes.