Many government shit is like that across the globe, mostly because government is incompetent in keeping up with tech and is underfunded/whatever/nobody cares about it.
It's not quite that clear-cut, though. Hard drives suck. SSDs suck. Flash drives suck. Floppies suck. Optical media REALLY sucks. It
all sucks.
The underlying problem is
we suck at data storage. There
is no reliable data storage. Just lots of it. Some of the smartest people on Earth have thrown their brains against this issue for decades, and the literal best-of-breed state-of-the-art approach for keeping data safe and preserving it even in the face of potential hardware failures is literally this: "maintain at least three online copies of every byte of data and keep each copy as physically far away from the others as possible, and the moment a storage unit fails, IMMEDIATELY SCRAMBLE to replicate the lost third copies it just took down with it somewhere else AS QUICKLY AS POSSIBLE, PRIORITIZED ABOVE ALL ELSE AND AT ALL COSTS."
Literally every large-scale data storage technology does this. Google's BigTable (and its successor GoogleFS), Hadoop, Ceph, S3, all of them do it. Ceph (to its credit) does support erasure coding pools ("raid5-ish parity") to reduce disk space wastage for replication, and the others experiment with it too, but the standard is still "keep multiple copies and separate each copy from the failure of other copies as much as possible." Fun fact: when you get an unexpected delay playing a video on Youtube, there's a
modest chance it's not actually them just being dicks about ad blockers or being deliberately slow -- they might actually have just lost a physical volume (or server) that had the copy of the video data they were going to serve to you, and their backend is stretching out further to grab another copy (at a lower priority than replicating it back to a "healthy" state again).
In terms of reliability, all storage media sucks. All of it. The most reliable we have are high-quality tape in well-protected, climate controlled environments, followed (theoretically) by those "M-Disc" physically-etched archival discs that claim 1,000 year survivability. The best we can do is just make a bunch of copies of everything and pray enough of it survives at one time to keep at least one copy alive.
So combine that challenge with your point of "underfunded," and yeah, it's an absolute pain in the ass to keep data "alive" in a reliable fashion. No matter how much money you throw at it, you can't get close to 100% reliable, so a lot of people just figure "eh, fuck it, throw it in a floppy and cross your fingers." I can't say I really blame them. Waiting for a hundred-node storage cluster to turn to a "healthy" state after a node burns up can be a real nail-biter.
