In IoT Large Environment Scenarios, complexity of networks can be too difficult to handle. When communication nodes (sensors, actuators) are too many, overhead caused by network protocols can saturate any short-range wireless technology that needs multi-hopping (e.g. Zigbee, 6lowPAN, 6TISCH, Bluetooth, WiFi, etc).
A (sub-optimal) solution taken in the past ten years is to split (divide) networks in sub-networks, moving the burden of their coordination (impera) to an upper level (e.g. in the cloud) where processing power, and storage capabilities are not a hard limited resource.
A more recent solution is to rely on Low Power Wide Area Networks (LoRaWAN, Sigfox, LTE-M, NB-IOT, etc) that allow to achieve long range communication (no need for multi-hopping topologies) at the cost of reduced data rates.
However, owing to limitations in terms of available bands, and the long ranges covered by such technologies, the density of IoT nodes that will be served might turn out to be unsatisfactory.
So, scalability (in terms of node density) of IoT applications remains a relevant issue. Whoever will find solutions to the issue, will have a say in the IoT business in the mid term.