Not if the filesystem itself can do atomic snapshots. The problem can happen if you try to copy the files or even the device when it's being written to. But if you create a snapshot and copy that, it would be consistent.

You can of course the up in a dirty state where some new transaction was started but not committed, but any non-toy database should be able to recover from that. (You'll lose the transaction of course)