In general no, but the provided example depends on parallel memory accesses at the cache level, so cache effects can indeed come into play with instruction-level parallelism. Did you just miss this detail in the article, or are you suggesting it's wrong?
Superscalar execution has nothing to do with caches. You could do it on an architecture with no caches at all.