Under high load, we had a very rare race condition in a common underlying framework that caused this. Before we detected and applied a fix, our logs indicate that it impacted about 12 users.
We immediately took down the servers immediately when we detected the issue and we put in a patch. Our engineering teams are currently working on it round the clock monitoring to ensure that it will not recur.
We understand that our entire business is based on the user’s trust, and we are taking it very seriously.