When there is an error (or keyboard interrupt), the subprocesses of the workers might still be running in the background. This can be problem for example in an interactive session, when workers still have their srun taking up the resources after the main program stopped. L2L should make sure to kill all the subprocesses when an error occurs.