It was the only game in town when communication technologies were primitive and expensive. When the cost per bit per hop is high, and the capital cost per direct link is also high, the optimal graph will probably be some sort of graph where the diameter (maximum degrees of separation between two nodes) is some sort of logarithmic function of the number of nodes. A hierarchical structure is generally the simplest way to optimize command and control under such conditions.

The cheaper it becomes to connect any two nodes in a system in ad-hoc ways, the more design freedom you have. Another key variable is simply the autonomy capability of each node the maximum sophistication of decisions it can take on. The higher that is, the less C2 type dynamics you need.

In historical terms, C2 for business was directly inherited from C2 for military. In the early days of corporations, companies like the East India Company were basically quasi-military organizations, and evolved along with military management technology into modern hierarchical forms.

As a general point, I tend to think people who turn organization design into a political football are being moronic. Structures arise from constraints, and every structure has room for exploitation and misuse. Polyannish visions of "networked, flat" organizations are no exception. They solve some problems better than hierarchical C2, under some conditions, but they are not immune to pathologies.

This is one reason documents like the Cluetrain Manifesto annoy me no end. The reflect a deep cluelessness about organizational realities.