Hi,
We have finally found what the problem was and solved it. We had the following flowable process with an event gateway:
This process is a common part called by other flowable processes. In some cases, timeout and answer was firing at the same time and both execution branches were continuing the process until the end of the calling flowable process, where Flowable Engine was trying, apparently, to commit both transactions and getting a deadlock in one of them. We have solved it by adding an async point (waiting point) at the end of the subprocess:
This async point (waiting point) forces to commit transaction before returning to calling flowable process. First execution branch to commit will return to calling flowable process and second one will throw a FlowableOptimisticLockingException and end.
Hope this help someone.
Thanks for your help, @wwitt
Greetings!