Ok so the risks (in terms of disk space) in this case scenario (assuming 1 VM on this datastore) is that the VM base disk will run out of free disk space.
Not the base disk, but the datastore may run out of disk space if a VM has active snapshots, or when you try to consolidate/delete snapshots with thin provisioned base disks.
Also as far as I understand you cannot resize the base disk if there is a snapshot applied.
Yes, that's corect. A snapshot .vmdk file contains metadata, based on the base disk's provisioned size at the time of the snapshot creation.
So one would want to avoid the situation in which the snapshot grows too much that it requires more free space to be consolidated/deleted but at the same time time you cannot expand the VM base disk because there is a snapshot. So basically you and end up with a chicken and egg scenario.
You are mixing up two things, free space in the base disk, and free disk space on the datastore. Resizing a virtual disk may only be required, if the guest OS requires more space. Deleting a snapshot however is transparent to a guest OS, and is only about the growth of the base .vmdk file on the datastore.
Example: If the thin provisioned base .vmdk file - at the time you create a snapshot - has a size of let's say 500GB, and you let the snapshot grow to ~1TB (assuming its additional data), then the .vmdk files will consume 1.5TB on the datasatore. Consolidating/deleting the snapshot at this point will not work anymore, because the data from the snapshot .vmdk file needs to be merged into the base .vmdk file (expanding t to 1.5TB) before the snapshot files can be deleted from the datastore. In this example the required disk space on the datastore would be ~2.5TB, which exceeds the 2TB datatore size. 1.5 TB for the base .vmdk file, and 1TB for the snapshot .vmdk file.
André