It seems like we need a better understanding of what nodes are, if they can be freely created and destroyed with no consequences, and/or if they get reused. It seems like either nodes must be freely creatable to a more or less nearly infinite amount, or there must be some method of perfect data storage that has an effectively unlimited capacity (a Turing tape?). Either way, things get kind of weird.
If nodes can be created and destroyed as needed, then there is no reason to apply different rules to an existing graph and mess up all its nodes, and it seems counter to the basic assumption of this entire system, which is causality (who knew the Merovingian was right?).
So the observer wouldn't constantly be in this pingponging state between existing and not and existing as a shoe or a goldfish or a bald guy with a chromebook and some spare time at work.
Now, if you assume that your node supply is limited, but you have infinite storage, then you can just kind of apply some kind of ur-verse principle of attention deficit disorder, where you progress a branchial construct through rulial (I want to call this Rulian space but for now I'll skip that bit) space, save it to Turing tape, and then load a different branchial construct and iterate it for a while, then go back to the other one (side note: is every character in Skyrim ending up as a stealthing archer an example of causal invariance??) and progress a bit further, etc. But this feels really inelegant and requires a Demon of some sort to hang around and constantly swap tapes for no good reason.
Since we have to assume at least one infinite resource for all for this to work we might as well assume that nodes are infinite and branchial spaces run in paralell in rulial (Rulian!) space, as that fits in with our models much better and also keeps with the central theme of all of it, that internally consistent causality is the underlying organizational principle of basically everything that we could recognize as any layer of the universe (it also hints that there might be a hypercomputer at the top, and that someone suggested to it the idea of causality and then asked it to compute every possible and impossible expression of that idea, which would look like it took time if you were inside the computation but which would be instantaneous from outside of it).