[Citation needed] -- Not saying this to be glib, but I'm just wondering what is an actual "huge open-source group". From what I can tell, most open source projects are heavily maintained by a single-digit number of people at any one time.

How huge is the largest open-source group though? My guess is that it's a order of magnitude smaller than the large companies being discussed in this conversation.

Communication overhead is MORE relevant than it ever has been. Sure, large open-source groups can collaborate on a project, but each person needs to know the operating principles and what element of solution is being held paramount. Regardless of inputs, mathematics can "only optimize 1 variable". Large groups can offer a larger pool of situational knowledge, which is valuable, but if individual A knows a certain element MUST be included, they now have to convince a great number of people that they're actually right. "There are a thousand ways to skin a cat", but if you put 1000 engineers in a forum and ask them to skin a cat, each will chose and champion 1 of the thousand ways (exaggerated for effect) and if left alone will complete it. The artistry is in developing a system to discern a solution that is above average, within the agreed upon timeline. A 4 person team might only brainstorm the 4-10 most obvious solutions, but the discernment process is orders of magnitude simpler.

I would guess this to be because of

1) self-motivation

2) people working towards proposals & being able to fork when unhappy.