Google Cloud regional or multi-regional storage is built around replicating several copies of blobs, intra-zone, between zones, and between regions. In theory, even a data center burning to the ground will result in very little data loss. Snapshots are stored like this, and some types of disks.
Snapshots automatically maintain at least one full with incrementals. This cannot be controlled with the API, not as far as I can tell. Usually I would want the option of all fulls for a simpler recovery, but if this can reliably do incremental recovery with an easy interface, that could work.
Yes, the storage backend losing a full snapshot could be catastrophic to point in time recovery. There may be data only in the full that is not in any incremental or the online disk. Unlikely given the architecture and their ops team, but would be a high impact event. At least snapshot create docs claim that you cannot delete the online disk in some failure modes:
If you attempt to create a snapshot from a zonal persistent disk and
the snapshotting process fails, you won't be able to delete the
original zonal persistent disk until you have cleaned up the failed
snapshot. This failsafe helps to prevent the accidental deletion of
source data in the event of an unsuccessful backup.
Simple fulls every time is not an option with the snapshot feature. Accomplishing that would imply adding your own additional backups. Perhaps use snapshots as the source, but archiving them somewhere. Additional cost and complexity, but not entirely relying on snapshots.
