It seems like for all the silliness and inefficiency that comes with a decentralized system ... the decentralized nature of US science research allowed for more "possibilities" and that paid off economically in spades.

Like speech, ideas require an open field with a lot of garbage to hit many home runs.

I expect every serious/successful researcher, artist, or other creative problem solver would agree that even within the ultimate centralization of work, all in one person, a low bar for exploration of ideas and potential solutions is helpful.

The problem terrain insights generated by many "failures" are what make resolving interesting trivial, silly and unlikely questions so helpful. They generate novel knowledge and new ways of thinking about things. They often point the way to useful but previously not envisioned work.

Edison and the long line of "failed" lightbulbs is a cliche, but still rich wisdom.

But 1000 Edisons working on 1000 highly different "light bulb" problems, sharing the seemingly random insights they each learn along the way, are going to make even faster progress -- often not in anticipated directions.

I think a lot of the decentralization also correlated up with a wide range of directions, with decisions to pursue activity made at much lower levels than happens today.