Can anyone speak to why those systems were considered more reliable? Are world agencies relying on this technology because of the inherent risk of rebuilding systems, or because of some attractive property of floppy disks?
It is similar to when my father's company switched from a mainframe that ran all their stuff and was written in the mid 80s to brand new off the shelf software. The replacement software is all about trying to make your needs fit into how the vendor's product works. His company unanimously believes the original software was significantly superior in terms of issues (it basically just ran for decades with almost no issues). The issue is the older tech keeps getting more niche, so you eventually kind of have to migrate sometimes.
Compare that to something like JavaScript running in Electron, where at basically every line there is a chance that something could return an error or take away control of the CPU messing up your time dependant code. Of course there are ways to handle that, but it's a lot more complicated compared to a CPU your code runs on exclusively. Or you just run into bugs with third party code that you need to work around, or update a library and find a patch version introduced a breaking change.
This is probably why PLCs are so popular in industry. You can make sure a tiny part of the system works correctly and is bug free. Then as long as the contract of how it communicates & interacts with other parts of the system is defined properly, the system as a whole will operate correctly.
Outside of the tech industry IT projects fail a lot, and they fail in ways that non-technical people do not/cannot/will not try to understand. From the perspective of your average middle ranking executive or civil servant any project involving computers is very likely to just randomly end up delivering nothing despite consuming tens of millions of dollars/euros. The phenomenon is utterly baffling to them and feels hopeless. Combine it with a rolling heads culture in which the organizational response to failure is firing / sidelining someone in order to show accountability (vs a rigorous post-mortem analysis), and you get a lot of organizational weirdness that boils down to everyone scrambling away from anything that involves technology modernization.
At some level the reason we have such a thing as a "tech company" or "startup" is because these are what we call organizations that aren't intimidated by software projects.
Government organizations are especially prone to ending up on obsolete 1980s/90s era tech because there's no market pressure or bonuses incentivizing them to improve things, because awareness of how often IT projects fail is much higher now than it was back then, and because civil servants are typically very focused on staying inside the civil service and making it to pensionable age. The bulk of their comp is weighted towards the end of their lives due to way more generous pensions than in the private sector so this gives them all an incentive to avoid any project that might have a concrete failure mode.
Within the tech industry as well. Replacing a running system tends to be high-risk for various reasons.
So I would guess this is more about the friction of moving away from a legacy system, as opposed to the technological benefits of floppies.
It actually takes a very very strong magnet in almost direct contact to alter a floppy.
Floppies are ridiculously unreliable. They were bad, even when they were brand new in a new drive. Failure rates of one in thirty or so wasn’t unusual. Installing large software packages was a gamble because disc 27 of 33 might have a bad sector and ruin your day.
Floppies written with one drive often wouldn’t work in another.
The drives themselves would often collect dust depending on how the fans inside the PC case were set up.
The drive head would often wear out the floppies, especially if they were used daily (e.g.: boot discs).
Keeping them in a hot car in summer could destroy them.
I learned about forward error correction (FEC) in desperation when I was in my early teens because it was the only robust way to retain data on floppies for any length of time. There were a whole bunch of compression formats that could split your data so that any n-of-m floppies could be used to recover your data in case a few failed.