It's something of a semantic difference whether worlds should be said to fuse in the MWI.
Consider a photon which passes through a two slit interference experiment. One way of describing it is to say that there are two worlds, one where the photon passes through one slit and one where it passes through the other. Then the worlds fuse when the photons hit the screen and this is the cause of the interference. Another way of describing it is that there is no split of worlds, that there is only one world in which the photon takes both paths (in some sense). Only when there is a measurement which collapses the wave function do we get a new world. In this view, if we put a detector at the slits which could tell which slit the photon passed through, the worlds would split when the detector was triggered. In one world the detector would show the photon going through one slit, and in the other world it would show the photon going through the other slit. It is the act of measurement, an irreversible act of amplification, which causes worlds to split in this definition. Both pictures are consistent and both use the concept of many worlds. It is just a matter of definition whether you want to say that parts of the state function which still are coherently entangled are separate worlds or not. If you use the second definition, though, I think it is correct to say that worlds do not fuse. Even a seemingly insignificant difference, if it occurs at the classical level, will produce a fundamentally different quantum wave function. And in fact, given the chaotic nature of the classical world (where the flap of a butterfly's wings causes next year's hurricane), it is likely that there are no insigificant differences. Hal [EMAIL PROTECTED]