Comment 23 for bug 1605089

@Han

Whatever triggered the update while the two workers were reading data will also trigger a backend update and re-read itself.

However, there is a problem where you can have a worker pause after it reads from the DB on the verify step, then another worker comes in, does an update, verifies the DB and returns. Then the first worker updates the backend. To solve that problem everyone would need to re-read until it converges.

However, a journal does not even help the above problem unless you serialize all updates and have some sort of election system so there is only one worker that can operate on a given object at a time.