A friend and I were discussing this, and came up with this picture:
The top graph is the instantaneous extraction of various flavors from a bean while it's steeping, while the bottom is the preference curve (which says I am a snob for green roasts, while he will drink any coffee-flavored swill). If you integrate the top graph, you will have the total extraction over the time period of your integration. This is the same thing as your steep time. One can also compare the two graphs directly to predict what kind of brewing someone might prefer. He seems to think I like underextracted (sour) coffee while he likes burnt coffee.
Anyway, the curves are obviously abstract and will change with different beans and roasts. My hypothesis for the Ethiopian Super Natural is that the bitter curve is quite a bit lower (perhaps due to the darker roast), and with a profile that is weighted lower, so the bitter hump occurs more to the left and it's less symmetric. The tasty curve could be biased more towards the right. Either or both could be true. That is, the bean could have the bitter curve in the picture, but have a tasty curve that peaks more to the right.
So for relatively low steep times, you'd tend to get more bitter than tasty components. Whereas if you steep longer, the higher extraction of the tasty particles could help offset the bitterness that you get with shorter extraction. The interesting question is (assuming this is true and we aren't just pulling things out of our butts) how one can shape or determine these curves for a bean so as to better customize a brew for a bean.
Hope this wasn't overthinking it.