At its current level of globalization, the US is at a point where it stands to benefit significantly in aggregate from further globalization [0]. The problem is that the benefits of globalization disproportionately accrue to the wealthy, both in aggregate and at the US's current level of globalization [1].
IMO, the best approach is to "grow the pie" by promoting globalization and free trade and then implement distributional policies domestically to ensure that no one gets screwed over. Most advocates of globalization would likely say the same. The issue is that in practice, that increase in globalization has occurred, but domestic policy has been regressive rather than progressive in terms of its impact on wealth inequality.
[0] https://www.imf.org/en/Publications/WP/Issues/2018/03/13/The..., page 23 of linked PDF
[1] Same source, page 26
The larger problem seems to be that our political institutions are captive to a hyper-polarized two party institution that lacks the degrees of freedom to provide legitimate representation to society's myriad facets and interests. Any dualistic system is naturally going to orient itself in opposition to its other half, to the marginalization of all other considerations. The key to effective electoral reform is finding ways to fundamentally change the processes from which structural bipartisanship emerges from. Practically speaking this means increasing political diversity through non-partisan primaries, ranked choice voting, proportional representation, and purging the entrenchment of the two national parties from election commissions and similar bodies.
Transforming the political court from a 2-sided chess match to a multiplayer game is a necessary condition for realizing the true expressive power of a democratic society over concentrated interests. Such conditions have been the key to the establishment and normalization of wealth equalizing institutions in many of the European democracies, for example.
Why does it "need to occur"?
For example eating local produce is better for the environment than eating stuff that has been transported for 1000s of miles, with the respective carbon footprint. And that's just an example from an environmental perspective.
We could grow just about anything indoors year-round, but it's much better for both the environment and my wallet to grow crops in places where we don't need to enclose, light and heat them.
Depends on the country -- and region of the country. There are countries that sport different climates and micro-climates. And there is stuff that thrives during the winter as well. And if it's really needed, we've had greenhouses for ages.
But that's part of the proposition: don't eat out of season stuff, even if it means changing the diet in an annual cycle (which, e.g. tons of Italians has no problem doing -- even if it means no, of much fewer tomatoes in winter).
While this isn't the point I was attempting to make in the sentence that you quoted, I also tend to think that increased globalization should occur, at least at the margin. But to your point, there may be specific sectors in which globalization is harmful in net, and externalities do need to be considered.
It's also likely that America could not have maintained its standard of living by walling itself off from this new technical development. The rest of the world would have continued to trade without America.
The economics profession has generally become one paid for its conclusions, not for its research. There are very few people working in mainstream, evidence-based economics anymore, and I don't think they're exceptions, they're just working for people whose personal interests happen to coincide with evidence-based theory. If the theory and the backers diverge, the economists will hopefully find new backers, because the backers will definitely find new economists who deliver the answers they want.
> actual science and engineering has made immense progress
I don't believe so. It's pretty much the same as economics. Most studies are wrong, and the review/reward system in science is all screwed up. We've generally been coasting on the invention of the transistor for a long time, and benefiting from their getting smaller and smaller. Once we've hit the hard limit with that, we can ride parallelism to squeeze a large amount of additional performance - hopefully enough to make a few more algorithms that have been around for 70 years have some pragmatic usefulness, but I think we're close to the end of what transistors are going to give us.
The best thing we can do for science is to get rid of the perverse incentives that cause bad science, eradicate accepted bad science, and mine the good science that will come to light after that uncovering. That might get us another good 100 years without a major discovery.