It seems that a hidden variables theory could reproduce Bell's experiments results, preserving locality, as long as it accepts that the superposition state(or at least a faux superposition) existed at $t=0$ of the creation of entangled pair. In this case, the question "where and how were the particles previous to the measurement?" makes sense, but only up to the moment they were defined to be opposites. It could be that, while movement over time for a entangled particle is also described by a wave function, its spin for each vector remains constant over time (when considering the same direction measured), only being uncertain at t=0. There's even a global vertical direction that could be responsible for this: gravity.
Bell, in fact, accepts that $\vec\lambda\cdot\vec p >0$ could properly describe the possibly states for a single particle in a local hidden variable theory. Why can't it be that when the entangled pair is created, this measurement is ran over in a particle, which defines every possible vector for this particle (respecting the probability), then the result is transferred and reversed in the other particle? That's all this local hidden theory would need to accept. In this case, the particles only communicate at $t=0$, and never again.