On August 27th 2026, I set a CPU limit of 32 for --qos=normal jobs.  I'd like to talk to users who have wider jobs, to see what we can do to make the queue flow more smoothly than it does with a big range of job sizes.  Based on one conversation, I created --qos=wide that allows 33+ CPUs, but limits the number of simultaneously-running jobs per user.

Please come talk to me if you have ideas.  If your need is more regular, I could imagine things like repeating scheduled reservations, to improve the accounting and reduce some of the dynamism that happens when batch scheduler is trying to drain nodes while the queue is being constantly reordered.  I want to enable your work!

For anyone searching, the error you get looks like

sbatch: error: QOSMaxCpuPerJobLimit
sbatch: error: Batch job submission failed: Job violates accounting/QOS policy (job submit limit, user's size and/or time limits)