Clustering:Leader and followers get to an inconsistent state when Leader transactions fail

Consider a scenario when there are concurrent writes on the same node in the Leader.
We would replicate these to the followers too.

When we would try to apply these to the state, one of those writes might fail (with an Optimistic lock exception) at the Leader end.

But since its serialized at the Follower end, these transactions might go through.

Since we dont have a way to rollback, this would result in an inconsistent state between the Leader and the Follower(s).