There is also a chance that the whole concept is completely confused; for example that "the average living copy of yourself in a thousand years" is a wrong way to predict your most likely personal experience in the future. Instead, the correct approach may be to talk about your typical person-moment in space-time, because there is no reason to privilege the future. I mean, you already know that many of your copies will not have person-moments in far future. And maybe your expected person-moment is plus or minus what you are experiencing now, precisely because you are not immortal.

This sounds interesting, but I don't quite get what you mean by saying that many copies won't have person-moments in the future or how this leads to non-immortality. Can you elaborate?

In general, I agree that estimating these probabilities is very difficult. I suppose the likeliest ways may, in any case, be orders of magnitude more likely than others; meaning that if QI works and, say, resurrection by a future AI or hypercivilization is the likeliest way to live for a hundred million years, the other alternatives may not matter much. But it's hard to say anything even remotely definite about it.