undesired removal of volumesnapshot CRDs

I have seen this in a few CI runs:

E   Failed: The following CRD(s) were recreated during upgrade (creationTimestamp changed), which destroys all existing custom resources of that type:
      - volumegroupsnapshotclasses.groupsnapshot.storage.k8s.io: install=2026-10-01T13:05:38Z  upgrade=2026-10-01T14:46:44Z
      - volumegroupsnapshotcontents.groupsnapshot.storage.k8s.io: install=2026-10-01T13:05:38Z  upgrade=2026-10-01T14:46:44Z
      - volumegroupsnapshots.groupsnapshot.storage.k8s.io: install=2026-10-01T13:05:38Z  upgrade=2026-10-01T14:46:44Z
      - volumesnapshotclasses.snapshot.storage.k8s.io: install=2026-10-01T13:05:38Z  upgrade=2026-10-01T14:46:44Z
      - volumesnapshotcontents.snapshot.storage.k8s.io: install=2026-10-01T13:05:38Z  upgrade=2026-10-01T14:46:44Z
      - volumesnapshots.snapshot.storage.k8s.io: install=2026-10-01T13:05:38Z  upgrade=2026-10-01T14:46:44Z

(https://gitlab.com/sylva-projects/sylva-core/-/jobs/16863475072 and https://gitlab.com/sylva-projects/sylva-core/-/jobs/16863475070 which were on a "☁️capo 🚀rke2 🎬sylva-upgrade 🛠️ha 🐧ubuntu" run)

The result of a CRD removal is a deletion of all the corresponding resources, which here would mean deleting VolumeSnapshots.

This is a very serious problem (release blocker).

(This is likely to be a side-effect of !8101 (merged))