1 pointThen, tbh, you need a better system for archiving that data. Like with backups (and I draw a distinction between archive and backup), you should be thinking "3-2-1" -- 3 copies on 2 different media with 1 offsite and, importantly, you should be rotating your archives on to new media more often you have been. High-resilience disk-based storage is relatively cheap and you should use that for your "primary" archive copy. Don't forget to check application versions etc -- you may be near the point where you'll need to re-write old data to new formats... I wouldn't know -- and I'm not sure I'd trust anything over Retrospect for recovering RS's proprietary format (tar tapes would be a different matter). Best to avoid any problems with "3-2-1", media rotation, regular checks that you can retrieve files, and so on.
1 pointDon't know about best, but what I'd try is: Rebuild to "tapeset1", starting with the first tape, expecting it to fail at the end of the tape as you've described Rebuild to "tapeset2", starting with the first tape, expecting it to fail... Take tape 1 out of the drive! Repair "tapeset2" and, when it says insert tape 1, mark tape 1 as missing and continue with tapes 2 and 3 You've now got the most data back that you can, albeit in two sets. You could then try and combine them by copying "tapeset1" to a new "diskset1", then "tapeset2" to that "diskset1" -- "diskset1" can then be moved to your newer machine for the conversion. Your choice as to whether "diskset1" is a File Set or Removable Media Set. You could, perhaps should, copy the two tape sets to two disk sets, convert them to the new format, then combine them -- it'll take longer but might be "safer". I'd also start to wonder about the chances of me ever needing to go back to data from 10+ years ago and whether it's worth all this extra work! That'd be a struggle between my innate laziness and my OCD, but you may have compliance issues to satisfy.
1 pointoslomike, Was the Tape Backup Set that contains the "bad" tape originally written with the Exabyte drive? I'm guessing that this may be so if the Backup Set was created by v4 or v5; the timeline of Retrospect Mac releases at the bottom of this Retrospect History article versus the Wikipedia timeline for the release of AIT-2 seems to show that a Backup Set written up through 1999 couldn't have been written on your AIT-2 drive. Therefore, if you still have it lying around, I suggest you try re-reading the tapes in that Backup Set using your old Exabyte drive. I'm not an expert, but differences in mechanical skew between drives can make tapes written on one drive hard to read on another—especially for special characters such as end-of-tape markers. Let me tell you a tape story: In 1969 I was working as a programmer at ITT Data Services in Paramus NJ, and my boss asked me to write a program to read "gapless" tapes on the IBM S/360 DOS computer at our satellite Lakehurst office and copy them to ordinary "gapped" tapes. The "gapless" tapes were being written by the U.S. Navy's Lakehurst Naval Air Station (that's where the Hindenburg dirigible caught fire in 1937), using a tape drive attached to a radio-telemetry receiver that had no ability to store received data while an inter-block gap was being written. LNAS could read these "gapless" tapes using a Control Data minicomputer connected to a special tape drive that could somehow "stop in less than 2 inches and navigate backwards to before the stop", but the minicomputer had such a pitifully-small main memory (4K 12-bit words) that the LNAS engineers (testing top-secret aircraft; don't ask 🙄 ) couldn't do any real analysis on it. The engineers wanted ITT to copy such tapes onto "gapped" ones that could be read by an ordinary tape drive on one of ITT's 512KB OS/360 IBM mainframe computers in Paramus. I wrote a tricky copy program in S/360 DOS Assembler, and it worked successfully on a test tape. However it turned out that the LNAS tape drive attached to the radio-telemetry receiver had a skewed read-write head, and the Control Data minicomputer's tape drive read-write head had been skewed to match. ITT wasn't willing to skew one of its satellite-office IBM tape drives to match, so my program never got used in production. See an audio expert for an explanation of tape skew.