I've observed a significant increase in the latency when creating jobs/pods at scale assigned to a Kueue's LocalQueue compared to an scenario not using Kueue.

In a compact cluster scenario using Kueue-operator 0.1.0, where we create either 500 replicas of pause pods or 500 replicas of a simple sleeping job.

In the case of pod ready latency, it goes from 2 seconds withouth Kueue to 40 seconds aprox when using Kueue

And in the case of jobs (with parallelism 1), StartTime latency increases from 0 seconds to 45 seconds aprox.