In this case it sounds like they just lost the build artifacts, not the actual assets. Although, when you read about how people have recovered the source for old programs it makes complete sense.
Without a central location where all the IP is stored, it becomes easy for older pieces to fall off the radar during company transitions, system upgrades, whatever. I suspect as people move away from large company wide SMB/NFS shares this is once again going to start being a common story. Particularly as people try to shrink their online storage footprint and move things to "archive" only. I've worked at places where I've had to repeating justify why i'm eating up a $LARGE_NUMBER GB for the last $SOME_NUMBER released versions of the product because the build process triggers a complete archive of the entire artifacts directories when a release build is specified.
Online storage be it, S3 buckets, or LTO backed NFS shares have real costs associated with them, and invariably IT comes along and starts asking why they have to upgrade their systems, or justify their OPEX numbers. Having a few TB of build artifacts that are rarely (if ever as the versions get older) referenced sits right up there with keeping previous employee's machine images, or home directories live.
The problem though is that putting things in an offline archive is just the first step in forgetting it exits. At that point your basically saying that in 10-15 years the only remaining copy of the data is going to be the git repo on some guys laptop sitting in his attic (the modern version of the floppies in shoeboxes).