Well, the good news is that it was fixable. Some data hiccups, but it's alive now. Apparently this has happened on some xenodes with ext3fs using 1k block size. A big thank you to caker for his help with this last night - above and beyond the call, as usual!