There's the natural limit and the theoretical limit. In theory, some future gene therapy may swap out your chromosomes for ones that don't decay (while somehow not causing negative side effects) while you're still an embryo or gamete, and do all sorts of other things to make you nearly biologically immortal. And in some even more distant future, maybe consciousness upload (with the "copy conundrum" somehow resolved) will be a real thing, in which case living for thousands or millions of years seems plausible.
That last case, "consciousness upload", means providing a new definition of "living". Nothing wrong with that, but worth noting, particularly since it's not at all clear what the new definition should be.
Yes, but it's not so discrete a difference. We change every second of our lives physically and mentally. We have ever increasing complexity of drug, surgical, and bionic therapies.
Running a simulation of a person would probably enable different kinds of change to be made, and some of those may be qualitatively very different from anything that has happened before. One also has to think about running multiple instances of a person, running a person in a non-real-time simulated environment, taking a checkpoint, restoring from back-up, and so on. At what point do you start to treat two instances of the same person as different people? If you stop running any instances of a person, have no particular plan to run one in the future, but you think you still have a snapshot in an archive somewhere, is that person still "living"?
