Ouch! I read the original post, thought: “Mesh,” thought about starting that discussion, decided against it, and now here we are. LOL!
OK, assume you live in a remote outpost in Alaska in a protected region and your only source of energy is solar. (Shipping costs mean a AA battery costs $15 and environmental conditions mean the batteries don’t work great anyway.)
Now design a sensor net.
Every message sent costs energy. So you don’t poll. You don’t send “message received” acknowledgements. You don’t check battery status more than once every few days.
You let your sensors sleep as much as possible. At the maximum practical interval, the sleepy sensor wakes up, checks delta against its maximum practical threshhold for the use case, if and only if the delta exceeds the threshhold send the briefest possible message, then the sensor goes back to sleep.
Nobody waits for anyone else, nobody sends extra acknowledgements, nobody cares if some other device is offline for minutes or even hours.
Energy use is maximized with very cheap devices. It’s a beautiful thing, from an engineering standpoint.
That’s mesh. Zwave and Zigbee as used in home automation. Cheap devices, very low energy draw, minimal human maintenance required.
What you don’t get is what WiFi babies grew up on: realtime monitoring and continuous status updates. But that comes at a cost. Higher energy costs, higher individual device costs, much more complexity all around.
You want that, you can have it. Apple’s HomeKit has it baked in. They’re assuming WiFi babies will never be able to make the adjustment to a mesh mindset. So their sensors will cost considerably more and their energy draw will be much higher. But you’ll have realtime monitoring.
If you have a use case that requires realtime monitoring and acknowledgement of every message, use WiFi devices. (Or zigbee in a star topology, which ST does not support.) But don’t deploy mesh and then try to turn it into something else. The results will not be pretty.
(My first commercial install was in a nuclear power plant. I understand there are use cases that require realtime monitoring. But you don’t solve them with mesh.)