In my case, the pre-script only makes available the directory that is to be cloud-synced. More precisely, in the pre-script I shut down the running service whose data is to be synced to ensure data consistency, take a snapshot, and then spin it back up to minimize downtime. I then run cloud sync on the snapshot directory in the special-purpose .zfs/snapshot/<snapshotname> subdirectory. (and delete the snapshot in the post-script).
There are some (dirty) ways to work around this, e.g. keep a snapshot alive to fake the directory so Cloud Sync doesn't complain, so I've marked the priority as low, but I still see it as a deficiency.
Please consider only evaluating whether the directory exists after running the pre-script (or add better native functionality for using snapshots as cloud sync data sources in PUSH mode!)