Two people can paint a house faster than one can. Honeybees work independently but pass messages to each other about conditions in the field. Many forms of concurrency , so obvious and natural in the real world, are actually pretty alien to the way we write programs today. It’s much easier to write a program assuming that there is one processor, one memory space, sequential execution and a God’s-eye view of the internal state. Language is a tool of thought as much as a means of expression, and the mindset embedded in the languages we use can get in the way. 
There’s a really interesting idea in this post. It’s that in the CAPs theorem, consistency usually get the boot, and we find that it’s not the end of the world. Updated state needs to propagate through the system, and hence, you get a speed-of-light effect like in the real world: light(information) has a finite speed, and you need light-cone diagrams to show you what you’re talking about.