That was my plan, along with using a distributed filesystem like glusterfs to replicate data across the network as backup. Unfortunately "time to go shopping" was ~1 month in on the second device (It was in a sub-optimal power situation...). A few hours, I had one machine with 7.5 TB of space and going on 4 months of uptime. ZFS has been historically BSD-only. Newer versions of the linux kernel are starting to support it, but I'm doubtful of the stability of a core module that hasn't undergone 10+ years of testing in the wild. It's got neat stuff to ensure data reliability on the drives, but if most people mount their storage over SMB (Windows shares) / AppleTalk, both of which negate the problem of ZFS support on OS X / Windows. Not sure about eSATA / USB...Whether or not you can dial into them when the host goes tits up I don't know; far as I'm concerned, if the host goes tits up it's time to go shopping anyway.
WTF is up with ZFS, by the way?