My views on the subject of a multiverse are:
1) The base level embedding system should have no net information.
2) The base level embedding system should have a dynamic.
The above seem to have consequences:
i) There can be no down select [limitation] on the number of worlds.
ii) There can be no down select on the properties of worlds.
Comments so far:
What is a world? In my view a world is just some sequence of temporary physical reality given to individual members of an infinite ensemble of preexisting packets of information I call kernels. Such members of the ensemble would be world kernels. A world kernel encodes a single state. A portion of some such kernels could be considered to be a memory [perhaps a false one] of past states. The dynamic of (2) gives a brief physical reality to world kernels in some sequence thereby producing a world.
iii) Each step of the dynamic must be inconsistent with its past.
Eventually the dynamic gives physical reality to world kernels in a sequence that has an evolution with respect to sub components [non isolated of course] within these kernels that seems to them consistent with some set of rules. There can be no down select on the types of rules.
I have posted my proposal for such a base level embedding system in the "An All/Nothing multiverse model" thread.