That's because "Worse is Better" provides revenue out of the gate. Design and performance doesn't sell, it just raises costs and causes churn. And so long as your features are bringing in more customers and revenue than you churn, nobody cares.
Yes, it will cause blowback later, when you have a feature-rich product that is unacceptably slow, or new features cost too much to be released on time, but that future may be years away, a timeframe you can't (or aren't allowed to) think about when investors are pushing hard for user and/or revenue growth.
Of course, those years later, it will probably be faster to pivot to a new bit of software written with a "Worse is Better" mindset and start the cycle again.
So, absolutely, work for simple interfaces. Absolutely, try to do the right thing rather than just creating a ball of mud. But the real world is a messy place, and your code needs to handle some of that mess in order to actually be useful rather than merely an entry in the museum of beautiful design.
Richard P. Gabriel in http://dreamsongs.com/WIB.html : "Simplicity -- the design must be simple, both in implementation and interface. It is more important for the interface to be simple than the implementation."
Ousterhout here is advocating The Right Thing, as opposed to Worse is Better. Unfortunately Worse is Better has become the dominant approach and it will be difficult to change that.