Think of it more like bigger building blocks rather than single use functions. If there is an issue with the pizza arriving burnt black at the customer you don't want to read through the logic for making the dough and adding toppings if the most likely cause is the oven.

Sure, you could add comment blocks to mark the sections. But you can't easily jump to that exact point. With function names you can easily skim over the unimportant calls, or go through a list of functions/methods in the file and jump there directly. With comments that's not a standard feature in IDEs.

Also that function does not scale well if you have more than 2 options of toppings. Maybe some are not compatible and logic needs to be added that you don't use pineapple and banana on the same pizza, for example.

But I understand your argument about following through multiple layers of abstractions. That's something that irritates me as well a bit, if you follow a function, that does nothing, but pass the same parameters through another function and return the result.

No guard clauses, or changes to the data, just 1:1 pass-through. And this them goes 2-3 levels deep before reaching real code again. Bonus of they all are in different files too.