The government could directly pay for domestic industrial capacity, if they want that. Recipients would get money for capacity, not for actual output. (And no tariffs would be necessary.)
Below is a sketch of a keyhole solution to this problem, that would be cheaper than tariffs and cause less disruption to the overall economy:
Recipients could prove that they have capacity either directly by just pointing to their output. Or if they want to claim standby capacity and want to get paid for that one as well, there would be randomised drills every so often where the government asks industry to produce a large quantity of the relevant items on short notice. Anyone who fails would forfeit a huge bond.
To give more details: suppose we want to make sure there's enough capacity to produce one million artillery shells per year. The government would auction off the capacity obligations to the lowest bidders. Companies that already produce the relevant items anyway would presumably have lower costs in fulfilling these obligations, thus they would be the primary bidders. But there might be some companies who operate purely on standby and don't keep a production running.
Being subject to a randomised production drill would be a huge expense, even if the government pays for the output, because of all the fixed costs you have in actually turning on unused capacity, even if only for a short time. But if the drills are truly randomised, insurance companies would really love to insure against them to spread out the load. Insurance companies could also insure against failing the challenge, and that would turn the insurers into private sector inspectors, because they'll want to make sure they don't undercharge companies for the risk of failing to meet their obligated capacity.
(See https://en.wikipedia.org/wiki/Catastrophe_bond for why randomised risks that are independent of the state of the general economy are so beloved by investors.)
If there's not much domestic production, then keeping standby capacity would be more expensive for the companies and thus for the government. Conversely, paying for standby capacity is a subsidy for the fixed costs of the relevant industries. (However it's a very targeted subsidies, because it goes to the lowest cost bidder. It's not sprinkled indiscriminately like a watering can.)
---
About defense related tariffs: putting tariffs on your closed allies like Canada and the rest of NATO doesn't make any sense. Steel produced in Canada is as good as steel produced in the US when it comes to defense. (Actually, it's a bit better in some respects, because Canadian steel workers don't have nearly as much political clout in the US as American workers and unions, thus the administration can't be blackmailed and coerced by them as easily.)
Even worse: the US administration allegedly wants to pivot to containing PR China and protecting Taiwan. Putting a 32% tariffs on Taiwan itself and 24% on key regional ally Japan is rather counterproductive.
But I think we both agree that Trump's tariffs were and are stupid, and we are discussing the best case for tariffs here.
So in the best case, it would still be silly to put defense-motivated tariffs on your closest allies.
---
Another addendum: the mechanism described in the first part might not work so well for software, because it's not standardised and has practically only fixed costs, no variable costs.
However, you could address that with other customised policy. Eg requiring open source software (especially when it's bought from overseas), and looking for specialised mechanisms for locals to demonstrate maintenance skills on standby or so.