I'm not sure what is more far fetched - many worlds or non-locality.
You're right that postulates such as the projection postulate complicate the theory. But it's not clear how to explain experimental results, like the stochastic nature of measurements or the Born rule, without it. I can make a theory as simple as you like if you don't require it to explain experimental results.
Again, I might be wrong, but I think the problem most people have with QM is that they think something must either exist or not exist. From that perspective it's easier to imagine that something exists in all possible states and that we are just observing one of them. However, even logically things can be in an indeterminate state.
For example, if I draw a box and then I draw another one that is twice as tall, the first box is short and the second box is tall. If I draw a third box that is 3 times as tall as the second one, then the first 2 boxes are short and the third one is tall. However, if I never drew the second or third box, then the first box is neither tall nor short. The concept doesn't actually make sense without the other boxes.
Similarly, imagine a universe with a single particle. What is its position? What is its momentum? Obviously these concepts have no meaning. As soon as I add another particle to the universe, then suddenly position and momentum have meaning. Now imagine that space is not a discrete map of particles with positions, but rather a set of particles which each have a probability of interacting with each other. This probability indicates their position.
In such a universe, if a particle does not interact with another particle, then it does not have a position. There is no meaning for position with respect to that particle. As soon as it interacts with another particle, then suddenly position has meaning. It is "created" in the same way that we "created" tall and short by drawing different sized boxes.
That is my, very possibly naive, way of thinking about it. Yes... I've read too much Taoist literature... ;-)
I would argue they did exist, and that the time travel event actually murdered them. I think of the universe as a computer, and any computation it does actually happens, actually exists in some sense. If the universe does a computation of my entire life, then deletes it, that life still happened. It "exists".
I think the same is true of the famous cat experiment. The computation of both possible states of the cat has to be done by the universe. It needs to compute how the cat reacts to the poison gas, it needs to compute the pain it experiences, etc. That's a lot different than a single particle like your example.
When you open the box, the universe may "collapse" to the second possibility, where the cat is still alive. But the first possibility still happened, and it may still be out there somewhere in an unobservable branch. In fact that model seems a lot more mathematically elegant, being completely deterministic and not having to deal with the weirdness of "collapse" events.
It may take exponential time to simulate the distribution of measured results given only the measurable inputs (which is what we do when we model experiments), but the universe itself knows the hidden variables and so can calculate the next state quite efficiently. So ontologically, a non-local hidden variables theory is much simpler than many worlds.
Copenhagen is unworkable ontollogically; it assumes some distinction between normal interactions and "measurements" which are poorly defined.
Do the others really assume that? Don't they still account for quantum erasure?