I’ll bite. Your first linked study does not say what you hope it does. In fact, it’s hopelessly lost and irrelevant. Here’s a passage that makes it clear “remote” is conflating multiple unrelated concepts and that the quantitative metrics are like Elon asking me to print out my code to determine “programs written per month”. A remote New Zealand vs Seattle worker is very different than a 1000 head site deeming two people organizationally distinct enough to plop in separate campuses.

We find that distant teammates impose negative externalities on the mentorship of teammates sitting together. These externalities can explain about a third of proxim- ity’s impact. Furthermore, before COVID-19, adding a new hire in another building reduces feedback among proximate teammates (who predate the new hire), while adding a new hire in the same building has no such impact. Teams’ attempts to ac- commodate distant teammates by, for example, moving in-person meetings online, have substantial negative externalities.7 Our findings indicate that additional mentorship has an opportunity cost: engi- neers who sit near all their teammates write fewer programs. Our difference-in- differences estimate suggests that proximity reduces programs written per month by 23 percent (p-value = 0.008), with similar effects on total lines of code and total files changed. The effects on output are present for both junior and senior engineers but are particularly pronounced for senior engineers, who do most of the mentoring.