[jandrewrogers]

On your first question, there are many areas of deep, hardcore technical domain expertise in things like databases, high-scale systems, parallel computing, AI, etc where much of the existing knowledge and recent advancements are poorly represented in public literature, conferences, etc due to layers of operational secrecy. What you would learn from the public literature or is shown in conferences is often quite misrepresentative of the state-of-the-art.

This becomes tribal knowledge. The way most people become experts in these domains is by working with or around people that are already experts, which requires being in an engineering environment where you are likely to come into contact with some. This type of expertise is far from evenly distributed.