Comment by heisenbit

Well, I expect people to understand http clients and if things don't work to be sufficiently knowledgeable to recognize when they have a performance problem and figure out why they have it. For that one needs language, library and networking skills which to a degree most developers have because they do it every day.

At issue however are niche skills. We are dealing with the long tail of a distribution and heuristics which work most of the time might not - the author mentions e.g. security. The way I look at this is risk i.e. security, bus factor, disruptions due to software moving from state "works and is not understood" to "broken and is not understood" and last but not least ability to predict behavior of this niche technology when it is going to be pushed into an larger project.