Recently we had in my work an outage due a race condition in production – a colleague mentioned that this is because of an not synchronized access to a HashMap in a multithreaded environment.
Well, I searched a bit around with my preferred search engine and found a very good description of what exactly is happening there: http://mailinator.blogspot.co.at/2009/06/beautiful-race-condition.html
Nothing else to add here, as this linked blog entry is already saying everything 