Are you using block storage? I.e. iSCSI or NFS for VMs? If it’s really only Nextcloud, an SLOG will be of absolutely no use.
EDIT: similarly a cache (L2ARC) vdev might actually degrade performance depending on the amount of memory the system has got. How much are you planning?
I’d consider a separate mirrored SSD pool and place the database on that. Perform hourly snapshots and replication to the HDD pool.
With 7 wide vdevs I would also consider RAIDZ3. Then again with the small drives you are using 4x 5 wide RAIDZ1 with two spares will give better performance with moderate risk.