You can easily end up with another sort of complexity, where the code is difficult to reason about because it is so decoupled and the methods are so short (but the call stack very long) and modules so small (but many). You can end up with unnecessary code that's difficult to spot because the logic is so spread out that the big picture is clouded.

Because lots of small functions increase the system complexity by creating interdependencies. You also don't have context in each function to understand the big picture and very small functions can be meaningless on its own. You have to jump through many functions just to figure out what is being done.

Code is a non-linear medium, depending on the path one takes thought the graph and the scale one operates at clean might mean a very different thing. Take a long look at the history of ideas, it might be obvious looking back that thing should be the way they are, but no one just knows them as if they are a prophet. Even if there are sign of something being ugly or a hack this does not give any guidance as what direction to take next.

The feasibility of functional programming in commercial settings.