I agree. Research is ideal for remote work, since you are effectively your own manager and by definition there is no specific vision that needs to be worked toward - you are discovering that vision yourself. It's the literal ivory tower. In my experience if you give non-R&D engineers and small teams total autonomy they will all end up building great solutions to problems that are slightly different than the problem you actually needed solved.
> I think figuring out the best way to use a team of 20, 100, 1000 engineers to accomplish business goals is just an incredibly hard problem...
Agreed, don't mean to downplay how difficult good management is. I think the key thing I've realized is that in a remote environment, bad management is so much more obvious and detrimental to the employee and organization. In office, if your manager doesn't communicate vision/objectives well you can always tap the shoulder of the guy next to you, or discretely pick your skip's brain in the break room. The rest of the organization can pick up the slack. But that's much harder remotely.