Hi, We’re having jobs crashing when scaling down (scaling up works well). What’s the best practice here? is there a way to signal not to scale down besides grace period?