Moving files across a network is potentially a dangerous thing to do - as far as the sending machine is concerned, the copy has completed when it is told by the receiving machine it has completed. If the receiving machine crashes before the data has been written out from cache (which is what is being simulated), you can get this issue. I'm not sure any OS would handle this any better and I would suspect Tiger will behave in exactly the same manner.
The real fault with this lies with the receiving machine: It shouldn't report success until the data has been written out to hard disk, but in the interests of performance many SMB implementations don't do this. Of course, the perception is the fault lies with the sending machine but IMO this is incorrect. Power or other failures always have the potential to cause data loss when hard disks are cached: It's a price paid for acceptable performance and for this reason I never move files but copy / verify / delete (later)