In my experience adding new features quickly is a loaded term. My guess is that Notch would know exactly how to add new features to Minecraft very quickly. He has intimate knowledge of the codebase. However, asking someone else to do so is where this ability to iterate will be lost.

I admire well written code, but a lot of times I see gold plating where is isn't really necessary, because "maybe one day we could add X". You still have to maintain the knowledge of where that easy addition could be plugged in.

Code is great, we can just throw parts of it away and refactor it to add new features. Which is possible if you follow best practices. ;)

Being able to hand of code to someone else is always going to be useful in any non trivial program. Very rarely are successful projects handled by one person in a vacuum.

I think the issue is in designing the initial structure you have to make decisions about which parts of the code will be important and how the relationships of objects etc will work. Sometimes if you have something that differs massively from your initial assumptions then you are in for an enormous refactoring job (as well as often a big data wrangling job if your app is already in production).

When I have done work for clients I have had fairly simple feature requests which I had not anticipated which have required fairly fundamental re-structuring of the codebase, if I had known these in advance I would have designed it differently.

Even if I had not known the specific changes in advance but I knew how much it would be likely to change then there are places where I would have also altered the design , but of course having said that there were areas where I spent allot of time creating code that allowed for flexibility where it was simply not required.