You might not care about records that are 100+ years old, but you might.

Space (both RAM and storage) is a premium on these systems and therefore much thought is given to what you place within any given row. As a result, no one stored unix timestamps as that would have required 10 characters more per timestamp per row, which would have had other ramifications on other parts of the code and/or memory. It would have required all stored records to be re-written somehow, which depending on the available online storage may have been an impossible task in and of itself. Not to mention how one would handle the recovery of any existing backups.

Due to these complexities, most organizations settled for code that shifted the meaning of the 2 digit year by some predetermined offset.