In recent years there seems to have been a growing discussion surrounding MWI's ontology. In the 2010 volume "Many Worlds?", Simon Saunders has a chapter dedicated to discussing whether the worlds in MWI branch or diverge.
The branching view states that before a measurement there is 1 observer in 1 world, after the measurement there is 2 observers in 2 worlds. The divergent view states that there were 2 observers and worlds all along, but they were identical up to the point of differentiation, so there were no way for you to tell which world you were really in.
Alastair Wilson has written several papers on this in the last years: http://alastairwilson.org/
Any MWI proponents got any opinion either way?