The capture process proved to be a challenging endeavor since some of the tapes involved were over 20 years old. The degradation of the media varied widely depending on the previous usage of the tape. We had initially expected that many of the tapes would have degraded to the point that they would only make it through the tape mechanism once before they either broke or had so much metal oxide flake off that they became unreadable. We also expected to find a great number of hard media errors, which the software would have to handle gracefully and efficiently, or we would never make any progress.
Much to our surprise, we have encountered far fewer of the failure modes than we had expected. Instead, our most common failure was much more counter-intuitive. We expected that the incremental dump tapes would be in the worst condition, since they had been overwritten repeatedly. In contrast to the incremental tapes, the full dumps were performed on brand new media and not accessed as frequently. It was the case that the incremental tapes had the highest incidence of read errors, but much to our surprise, the quality of the tapes improved after we ran through the entire read process a few times. We speculate that we were removing loose oxide from the back side of the tape, which had adhered to it from the next turn on the tape spool. Initially, this loose oxide interferes with reading, but after we removed it, the tape became more readable.