The thing is, you don't actually need to postulate that the cat goes "poof" at the point of observation. Suppose this didn't happen. Then the state of the world before observation is this:
(Normalization dropped.) This isn't weird. What would be weird is if these two states could interfere with each other.
I did some work a few years back which argues that interference between these states is virtually impossible - they live too far away from each other in configuration space.
Ultimately the argument goes all the way back to David Bohm in the 50's. The novelty of what I did is cooking up a plausible experiment for which the calculations are tractable.
Cool. How do you deal with the double slit experiments if there is no interference? Aren't you back to a fully deterministic worldview, from the observer's perspective?
There is interference - my model is merely the many body schrodinger equation, after reduction to a two-particle NLS.
The key observation is that a measurement apparatus involves N particles, each moving a distance O(1). Thus, the distance (in configuration space) between both wavepackets (i.e., |cat dead, observer sad> and |cat alive, observer happy>) is O(sqrt(N)).
Because the two wavepackets are so far apart, they can't interfere. This is true with or without the observer, since the cat still has O(10^23) particles.
On the other hand, if there were no observer, and the system were merely an electron, the distance between wavepackets would be O(1) and interference would be possible.
I did some work a few years back which argues that interference between these states is virtually impossible - they live too far away from each other in configuration space.
http://cims.nyu.edu/~stucchio/pubs/slides/wpi_wave_collapse....
Ultimately the argument goes all the way back to David Bohm in the 50's. The novelty of what I did is cooking up a plausible experiment for which the calculations are tractable.