Hi verto and welcome to the community,
The issue is that if enough blocks within the guest change and the thinly provisioned VMDK's need to expand then the VMs will pause as further writes will not be able to happen to this datastore. At that point you will either need to expand the datastore or free up space on the datastore.
Over committing disk this way requires careful management of the VMDK expansion and you need to constantly monitor the freespace on the datastore to ensure continued service.
Kind regards.