Personally I've never seen any web site or app that could be called "as effective/perfect as it's really ever going to get", frankly, not even close.
I browsed Chen's article and he talks about how you could always still do small incremental updates but the impact of those changes, the overall improvement they make, decreases - you put more and more effort to get less and less benefit is what I gathered he's talking about. That makes sense to me.
In filmmaking there's an old saying: "movies are never done, they are simply abandoned" - at some point you just have to walk away and begin work on another project - and maybe that's the answer to your question: you know you've reached the "local maximum" when more value could be delivered to your users by giving them another product.
Or I could be completely wrong!