Hrm… My experiments show that a job is executed three times, not four, and then an incident is created. Hence the number of executions is three, but the number of retries is two.
Is this me who does something wrong or does the camunda docs use a not quite precise word?
The number or retries is actually the overall number of a job execution until an incident is created. So the documentation might be a bit misleading here.
If you want to force zero retries you can pass “R0/PT1M” to the “Retry Time Cycle” field in the modeler (it says “retry zero times for a period of time of one minute”). This way an incident will be created right after the first run failed.