My interpretation of the double-slit experiment has been that the presence of physical matters in a certain configuration imposes some sort of probabilistic "reality field" that propagates at light-speed. The superposition of these "reality fields" at a certain point in space(-time) determines the probability of the particles' behaviour at that point, not unlike how support vector machines with a RBF kernel work.

This interpretation has the benefit (I think) of bypassing the problems with the "observers" --- any apparatus that is able to discern the path a particle had taken just so happens to impose a field such that, when superposed with the original "reality field" the double-slit imposes, the mixture field no longer exhibits the interference pattern.

Of course I am not a physicist by training, so the above could be completely nonsense (or it could have already been an established theory).