This release improves scheduling accuracy for newly created nodes with device-plugin-managed resources such as GPUs. During the brief registration delay, catalog-declared extended capacity is considered so workloads do not trigger unnecessary duplicate machine requests. Once a device plugin reports allocatable capacity, its reported value remains authoritative. ## Changes ### Features - Prevents duplicate machine provisioning for pods requesting catalog-declared extended resources while a new node's device plugin is still registering. (!48) ## Contributors - @d.hauck **Full changelog:** https://gitlab.com/kateops/k8ops/-/compare/fec53c514e9cf613c911e7f7757e37a1820bbcd4...6c5fcba8df631089647ca30fe406e7cd6cc458d8 Released from main (fec53c51..6c5fcba8, !48) by KateopsAI: merged and published automatically (every merge request carries release::autorelease).