I see now where the problem is. It's basically a contention between two or more DB transactions that update state of the same workflow "simulteneously". This should have been solved by acquiring a lock on workflow execution object so that all changes are atomic. Seems like this simple locking mechanism doesn't work properly. I'll investigate why. Keep working on it..
I see now where the problem is. It's basically a contention between two or more DB transactions that update state of the same workflow "simulteneously". This should have been solved by acquiring a lock on workflow execution object so that all changes are atomic. Seems like this simple locking mechanism doesn't work properly. I'll investigate why. Keep working on it..