Hello together,
there is no "right" or "wrong" answer to this questions imo. Both concepts have their two sides.
Personally, I prefer to go the multisensor way (and also use RS485, which limits the numbers of sensors to 32 per gateway with standard interface boards). In addition to that I also like the idea to let the arduinos decide (some) things locally or make them configurable to some extend. So things may not become completely out of controll in case there is (temporary) no controller.
Advantages:
Less parts, especially power adopters/batteries,
less "optic irritation"
better use of arduino potential (use of more than 40% of the memory, not sleeping all the time)
Less RF, means also lower risk of RF conflicts (or best: no RF at all => RS485)
in case of e.g. failure or power off, I will most likely notice this earlier because of the bigger "footprint"
Disadvantages:
its (sometimes) hard work to get things working (but also fun and satisfaction )
higher risk of messing things up because of bad code (I am just a beginner in programming. This is more related to distributed logic as mentioned above; just combining some simple examples isn't rocket science...)
in case of e.g. failure or power off, a multisensor has the bigger "footprint"
So it's up to you which way to go (or to which extent, to be more precise).
