It is unlikely that society will ever neatly divide into "haves" and "have nots"- be suspicious of sharp divisions. There should be lots of boundary cases, and probably a smooth gradient.

The primary question for answering these sorts of questions, I think, is whether modification to existing agents is more or less effective than modification to new agents. It seems more likely to me that genetic engineering can radically increase human lifespans than interventions in already developed humans- and if that's the case, then there won't really be immortal elites because the younger generation will always have rosier prospects than the older generation (unless a cap is reached, of course). Currently, we would expect a 200 year old programmer to run circles around a 20 year old programmer, because experience is really valuable. But if the 20 year old has 10 generations of intelligence boosts that the 200 year old can't get, then the 20 year old is probably going to win- especially if one of those boosts is easier experience transfer!