If there is another country (USA) which has relatively significantly diminished supply chains and manufacturing capacity and therefore has an even greater national security risk, why is that not applicable to be discussed? If, further, there is an opportunity and process by which that country (USA) could close that gap, why is it not applicable to be discussed?
You don't need to keep spare capacity for everything, but you need to have local capacity (not necessarily spare) for building custom manufacturing machines and staffing them, so that you can divert it to manufacture arbitrary physical goods if needed. Some countries can do that, most don't - but some of these could and should build that capacity.
For N95 masks as the article points out, yes, no one has the melt-blown fabric manufacturing machines that are a precursor.
However, if everyone around you wears surgical masks, then everyone is blocking ingesting about about 80% of particulate matter [1]. For an emergency like the pandemic, we're also interested in the masks blocking what wearers breathe out to surgical standards, what the masks were designed for. That's a "not letting perfect be the enemy of good" level of "good enough" if your alternative is letting front line healthcare staff go without melt-blown fabric filtering PPE because everyone is scrambling for P95 or P100 masks.
The US military establishment after the pandemic is controlled will hopefully be given a hand in determining precursor tech tree bottlenecks like the OP article pointed out, and have Congressional authorization to perform economic disaster preparedness within its scope. Stockpile and maintain sufficient operational skills alongside key equipment, so our disaster preparedness is better than "duck and cover" (some interesting questions behind that door).
Some modeling suggests we need a minimum 3 month lockdown in the US to cut deaths in half [2], 18 months (or however long it takes for critical mass of population to develop immunity or vaccinated immunity) to approach a semblance of status ante quo death rate. I wish they would put their modeling source and data onto Github (or Wolfram Alpha?), so others can work with the model (I'm interested in its backtest results from previous epidemics). If we lockdown for even 3 months, then that's GFC effects; 18 months is Great(er) Depression.
[1] https://fastlifehacks.com/n95-vs-ffp/
[2] https://www.imperial.ac.uk/media/imperial-college/medicine/s...