Naginator uses a delay before it lets Jenkins execute the job again.

But at that time the job is effectively placed at the active end of the build queue.

The causes other jobs to be delayed when they are still there.

In a MVN project, that is sometimes critical, as the cause of the failure may not be in the job itself, but in one of its dependencies. Thus the job is merely waiting for the dependent job to be rebuild, before it itself can be rebuild successfully.

Unless the delay time is huge, there is no guarantee that the dependent component is rebuilt during the Naginator delay time. When that happens Naginator causes an endless loop in the build process. This is because time is spent on other failing jobs. But before the job that provides the solution has been started, the retry of the original kicks in.

An obvious solution is to make the Naginator delay time so large that all modules of the project can be built in the meantime. In my case that would be more than 3 hours. Which is a bummer since most of the time just a few minutes is enough.