Your disk's b*ggered. Switch it off for 24 hours, then get as much off as you can, print off the error messages, and then take it back under warranty!
No you're on the wrong tree here (and I wish it was my box). Smartmontools doesn't find any disk errors and all disks have 0 reallocated sectors. The problem is somewhere along the IDE bus. There's a third 80GB disk in hdc, so I swapped hdb and hdc, and the problem is still on hdb whereas the previously troublesome hdb works fine as hdc.
I suspect a bad interaction between hda and whatever is in hdb (unless someone has a better suggestion). Thrashing hdb read/write with dd doesn't show it but unpacking a 400MB kills it. Don't ask me why... If the problem was a bad cable or caddy, thrashing gigabytes with dd should show it.
It could be something else. LM sensors show reasonable values? What about the HDD temperatures from smartctl?
Some more specifics on the motherboard may be useful as it could be something else. Also, there was that dodgy capacitor issue with motherboards around that pedigree (266/333 motherboards). We have one in the office that just plain locks up under heavy load, or a few days, or just randomly.
As solution I ordered a PCI IDE card to give each disk its own cable. Fingers crossed...
Just remember that it may reorder your hard drives (I know mine did). I believe you can pass a kernel boot parameter to stop it.
Regards
Daniel
