At work, we use AIT-2 tapes for backups (both the 36GB and 50GB type). This is all a bit unconfirmed, but we've had a few problems with read/write errors. It seems an "erase" function on the tape resolves the issue.
<--break-->
We have a 12 slot robot tape drive, which is obviously constantly swapping tapes around to get the normal (and majority of) backups done. We've had the odd tape problem, but nothing too serious to worry about.
Elsewhere, we have a single AIT drive which we use for Unix OS backups (data being covered by the other drive). It just uses ufsdump to backup the local disks of a handful of Sun machines. On this tape drive, tapes have steadily been problematic, with ufsdump reporting tape i/o errors (a long way into the tape). I assumed the tapes were getting old, so bought new ones (which work fine, incidentally).
However, it seems that running the "mt" command with "erase" (ie. mt -f /dev/rmt/0cbn erase") prolongs the life of the tapes. In fact, even the "erase" reported i/o errors... very curious... But, doing this seems to have resolved the errors seen, with backups now continuting right to the end of the tape without problems.
I suspect the truth of the matter is that the tape is a bit old (it's about 2.5-3 years old), but it hasn't actually been used that much (we have/had a five tape rotation, with weekly or fortnightly tape swaps - hardly a great deal of tape spinning!). I guess (like floppies) you have to degauss tapes from time to time so that the machine can properly format it again. I suspect this is into law of diminishing returns, in that I expect these tapes to be a problem once again, but until then, it's some use and saves a bit of cash replacing the tapes (although I've got some new ones which are the 50GB type, because the little ones are a bit small).
Thinking about it, the old tapes my peers used when they were young used to have to be degaussed every so often - so much so there was even a machine to do just that!