Who said I was using windows?
Ah but the other file system should be similar...
Anyway, the drive looks like it's on its way out. CRC errors occur trying to copy some files which is generally not a good sign. Once I have most of the data off I will do a scan+remap with MHDD and see what gives.
Yeah, wasn't arguing that the drive might fail, was just pointing out that usually data doesn't just disappear in my experience

Would rather be unreadable... But then again, I haven't experienced that sort of failure in Linux or BSD as yet (because I only use Linux now and then and BSD on my NAS which has RAID-z) and I don't use Mac.
Take2 rejected my "returns request" today. According to their suppliers, there is nothing wrong with my hard drive (Seagate Barracuda 7200.12 500GB ST3500418AS). The hard drive itself (S.M.A.R.T.) doesn't agree, though!
First, it reported 67 reallocated sectors and the following day 72. Of course it still works, but for how long? I just hope it fails "properly" before the so-called 3 year warranty expires.
Ouch. I see you are a Linux user, if the drive is on it's way out you can easy cause it to relocate more bad sectors (therefore quicker failure).
Just do: sudo dd if=/dev/zero of=/dev/sda bs=4096
Naturally where /dev/sda is the drive in question (eg. /dev/hda, /deb/sdb, drive designations can be found doing a sudo fdisk -l). You probably know all those commands anyway but yeah just posted them anyway.
It'll fill up the drive with zeros (all data lost) which I've found exposes bad sectors very quickly, also causes all bad sectors to be relocated.
It is normal for a drive to relocate a certain number of sectors tho, if there was a head crash however a large amount of sectors will be relocated in a short period of time and the drive will fail in no time. If it isn't a head crash could just be a few bad sectors that were there for a long time but were only detected because of large file accesses, format, etc.
Don't think they'll authorize a swap out without an actual S.M.A.R.T fault or other significant failure.