I like the bad block list that you suggested. What I am just curious is why everybody relies on the bad block relocation feature of modern drives and is reluctant to implement a real bad block relocation feature in mdadm or LVM2 (which from my point of view would have many advantages)?
I just had the experience that in a RAID5 array with 4 old 250 GB drives one had a bad block and was actually running out of spare blocks to reallocate that bad block. So long so good, since mdadm could not do a software bad block relocation I changed the drive with a newly bought one. When changing the drive I must somehow have touched one of the other hitherto functioning drives a little bit to roughly so when I had put in the new drive and tried to start reassembling the degraded array, that failed, too, since another drive also then showed a bad block and failed as well. Afterwards getting my data back was a hack of an effort with two drives failing just because of two stupid bad blocks.
I am currently considering reverting to evms with debian etch since this has a bad block relocation feature. But honestly I would be happy to use mdadm with a real bad block relocation feature since evms is not maintained anymore.