-snip-
In my semantics class, one of the talking points was that there is actually
no redundancy in natural languages (in a synchronous perspective, as in, in the language's current state at a given time), or at least, there is a tendency toward elimination of redundancy.
Now, that may seem to contradict my previous bit about redundancy being everywhere, but you seemed to interpret redundancy as having multiple words referring to the same thing, but differing by degrees of meaning, and I just went with it. What I mean here is that there is a tendency toward reduction of
true redundancy - words that refer to the same thing and have no difference in degrees of meaning, attached value judgment, stylistics, etc. As in, with every usual (I mean "which is in use", dunno if this is the English term) pair of words with seemingly similar meanings, there's going to be at least one context where one is acceptable from a native speaker's perspective, while the other is not. So, there's no word pairs that have a completely identical range of acceptable contexts. If such a pair was to appear, one of the words in it would fall out of use, is what I've been taught. It seemed quite convincing.
Redundancy as you're describing it (the multiple shades of meaning thing) is just an universal for any natural language, and as such I have no strong feelings towards it one way or another. It's just how things are. I am, however, a second-year ling student, so I'm just approaching it from a less artistic perspective because that's what I'm supposed to do... I guess. I haven't figured it out completely yet.