I was able to verify that this feature worked between a Windows 8 beta client machine and a Windows Server 8 beta server. This is a good feature for branch offices, or even hosted applications requiring file access whose traffic travels over larger wide-area networks or the Internet.
The beta includes some welcome improvements to offline files, including an awareness of employees' increasingly mobile lifestyles. For one, users can be designated as always offline, so that files are automatically cached to them and all work local to the user is done from that cache. The cache's updates are then filtered back to the host server or share (or distributed file system configuration) every two hours.
One caveat: The Previous Versions feature, which allows you to recover an earlier version of a document saved via shadow copies, is only available when the connection to the server hosting those shadow copies is live. I was able to make a user in my environment permanently offline by enabling the "Remove 'Work Offline' command" GPO; when I logged on as that user on a Windows 8 client machine, I was not able to remove myself from offline status. Things worked as expected from there.
The operating system understands the type of connection being used; if it's able to detect that expensive 3G and 4G cellular connections are active, it will cease the synchronization process until a cheaper connection is available.
The operating system also will include the ability for Offline Files synchronization to understand the type of connection being used; if it's able to detect that expensive 3G and 4G cellular connections are active, it will cease the synchronization process until a cheaper connection is available. (I wasn't able to test this feature.) This setting is available through Group Policy.
ReFS, the resilient file system
Microsoft made headlines in early 2012 by announcing its new Resilient File System, or ReFS. ReFS is designed as an evolution of NTFS with a focus on availability and integrity. ReFS writes to different locations on disk in an atomic, or transactional, fashion, improving data resiliency in the event of a power failure during a write.
For example, if a write operation was being committed and the electric power went out, writing in an atomic fashion would mean the write completes fully and no old data -- the data being overwritten -- would be left intact after the write. If you were to write to a disk non-atomically, during the power failure it's possible for the newly written data to be mixed, or overlaid, with the old data and the who thing ends up being just garbage.
ReFS also includes the new "integrity streams" feature that uses checksums and real-time allocations to protect the sequencing and access of both system and user data.