Adding manpower doesn't make product/shipping grow linearly. At a certain point the majority of work is coordination costs, but is mostly requisite to eke out a wide range of coordinated and successful products.
i think you underestimate the massive amount of backend and internal optimization that's required. Sure, the MVP of Twitter is a few thousand lines of code, but think about ad buying, ad display, abuse detection, customer service, content management, APIs, mobile, mobile web, etc., and that's just on the engineering side of things. For every engineer you'll need business folks, PMs, designers, operations, etc.
Hell, Twitter is a great example. We know what Twitter looks like without enough staff: an unreliable MVP that fail-whales all the time. Twitter acquired the staff and capital to scale up and they became far more stable.
The economies of scale large computer systems offer are counter-weighted by the n-squared complexity of larger systems.
