This issue was introduced in jBPM 3.2.10.

Because I don't understand why the the number of retries is hardcoded (and no use is made of the already injected value via jbpm.job.retries and the ObjectFactory), I'm digging deeper in order to make sure it wasn't just a typo, or in other words, that the programmer introduced the bug by accident.

(Otherwise, I would have simply fixed the code quickly and attached the commit to JBPM-3216).

In all other classes, (DispatcherThread, JobExecutorThread, Job), use is made of the jbpm.job.retries parameter.