Neighbour discovery is crucial for mobile ad-hoc networks and especially in Internet of Things (IoT) networks which are distributed by nature and built on mobile, communicating, and mostly battery-powered autonomous IoT devices. The challenge of efficient neighbour discovery in such architecture is to achieve both low latency and high energy efficiency discovery process, which have been shown to be major issues in the scientific community. In this paper, we contribute to the resolution of this problem applied on Bluetooth Low Energy (BLE) personal area network. A new neighbour discovery protocol called Snowball is proposed, which leverages the advertising and scanning features of BLE beaconing mode. Snowball provides new delay parameters upon the BLE stack to achieve a stable duty cycle schedule in the neighbourhood, which guarantee an efficient and low energy discovery process. The performance of Snowball is discussed and evaluated by simulation using Netlogo, and also by a practical experimentation using ESP32-based IoT devices.