Fixing this issue will allow non-execjob_end and execution hook to accumulate the job resources_used values across multiple nodes.