DV can use snapshots of metadata to define virtual databases, but doesn't take snapshots of data other than through the materialization options.

DV doesn't preclude you from reading from one source and writing to another.   And that task could be scheduled such that it passes a parameter on what it retrieves, such that it pulls only what deems a snapshot.    Another option on this could be to write the snapshot to a different source based on partitioning needs.   Using the multi-source option, you could control that source dynamically.     Just something else to consider.