Humans, and other large animals, are prone to infectious diseases. Yet large animals aren't extinct, which means that infectious diseases seldom kill off an entire species. Which leads me to suppose that for any species and any infectious agent, some members of that species will probably be immune to that agent.
It has been hypothesized that the main advantage of sexual reproduction -- which causes every member of a species to be genetically different -- is resistance to microbes and other parasites.
The main advantage of sexual reproduction is faster evolution. It makes evolution like a huge parallel computer instead of like lots of small, very slow serial computers.
If you examine asexual reproduction: you get a "tree of evolution". Each node in that tree is slightly "more evolved" than its parent on average (natural selection prunes the devolved ones more than the evolved ones).
However, each branch in the tree has to evolve every single feature independently from its sibling branches. You only inherit evolutionary enhancements of your own ancestors.
Assuming a tree of depth N: each node in the tree has only O(N) ancestors, and thus, about O(N) evolution experienced in each node.
With sexual reproduction, you also recombine the evolutionary advancements of two ancestors into new nodes. It is no longer a tree, but a DAG of nodes (Directed Acyclic Graph). Each node now has O(2^depth) ancestors. Evolution can recombine features that took a long time to develop into a new baseline, and evolve from there.
This is an exponential increase in evolution speed -- and I believe that is so incredibly huge of an advancement, that I predict that even species we believe are incapable of sexual reproduction, are probably sexually reproducing (e.g: exchanging DNA in some way) because it is too huge of an advantage to give up.
Lastly, I find that it is interesting that there's a similarity between sexual-based evolution process to one of the frameworks we use for parallel computing: "Map/Reduce". Take a single node in the tree, and just "Map" (mutate & select) over it multiple times, you get slightly better nodes. Now combining the mutated & selected nodes together into descendants is much like "Reduce". "Map" is not a useful parallel computation, but "Map & Reduce" is.
That's the basic idea but single celled organisms can and do directly swap DNA segments at any point in their life cycle. Unfortunately, multicellular can only meaningfully swap DNA when they are at the single cell stage. So sex is simply the maximum DNA exchange possible while multicellular life is at the single cell stage.
The disadvantage to sex is you need vary similar organisms so your N^2 ancestors becomes (species population size) * number of generations. Where direct DNA exchange can cross DNA from far less compatible organisms (~total life population size)
It has been hypothesized that the main advantage of sexual reproduction -- which causes every member of a species to be genetically different -- is resistance to microbes and other parasites.