Then add “earthquake” to the list, or “domestic terrorists or foreign country bombing the building”. Steelman the argument. The point isn’t “just fire and water specifically”, we’re not playing Pokémon.

We have several historic examples of records being lost in disasters, and way more recent than 100 years ago.

https://en.wikipedia.org/wiki/National_Personnel_Records_Cen...

It makes no difference that we could’ve prevented that with better building construction. We didn’t, and hindsight does not bring the records back. We should plan for the world we want but cannot ignore the world we have.

I’m not defending digital as always better or criticising physical. Like I said, different tradeoffs, meaning there are advantages and disadvantages to both, there’s no solution which is better in all situations.

I stuck to the threats you mentioned. Paper in a file room is more slightly more quake-resistant and bomb-resistant than digital. But slower to move to safety if the threat is large volcanic eruptions.

I am not saying that paper is magically perfect. Nor better in every situation. I am saying that paper is far easier (than digital) to do well for use cases like a national records collection. "Correctly" may include off-site backups - whether or not your threat model includes massive earthquakes, volcanoes, bombs, special forces, EMP weapons, biological agents, civil war, radioactive fallout, or enemy occupation. Or "Management wouldn't pay for a done-right facility".

As I noted in another comment, the largest downside to paper (within such use cases), is that it is far more difficult to get political support for old-fashioned stuff that just works, compared to anything that can be sold as cool/new/high-tech. Especially when the taxpayer-funded revenue streams from selling/installing/supporting the tech create incentives clearly contrary to the taxpaper's long-term interests.