A lot of it depends on scale. At facebook's scale (user base and revenue), it's worth paying 10 engineers to make little tweaks in a narrow area of the product because that little polish, while not worth doing for a smaller company, may bring in an additional 10s or 100s of millions in revenue. Similarly for tooling. Most shops can and should use stuff off the shelf. But when you have thousands or 10s of thousands of engineers, making all of their lives even a little bit easier is worth putting together a team.