Not sure what you mean by this. At least in a biological setting, natural selection selects for `adaptive' features, i.e., those which best solve some problem that an organism has. (OK, it's more complicated than that, but let's take that as a working definition.) It does not select `popular' features. Rather, features become popular because they are adaptive and hence selected for.
Now, programming language evolution may not always work like biological evolution, and in both cases there may be other selection pressures at work. But I don't see a reason to think that, insofar as programming language evolution is like biological evolution, the `natural selection' of language features by the community would favor what's arbitrarily popular, instead of what's best at solving a common problem.
That's because the author isn't saying that. The point is more like this: Clojure is a series of new mutations, and they haven't been around long enough yet to tell whether any of them will survive natural selection. We can't tell Clojure's good ideas from its merely new ideas at this stage; we have to wait for evolution to take place. That doesn't mean Clojure is doing something wrong (after all, "Everyone should be writing new Lisps!"). I think the author is just saying it's not worth getting too excited about until we know what it's doing right.