Bluetooth Low Energy (BLE) is envisioned as one of the most prominent technologies for the Internet of Things. The recent release of BLE mesh has attracted interest from both the industrial and academic communities, with new proposals for optimizing its performance being published on a regular basis. However, the majority of these publications rely on analysis and simulations, potentially neglecting the effect of real-life environments and hardware limitations. Moreover, no framework is available to experimentally compare the performance of different mesh protocols on top of BLE. This article presents the design and implementation of a low-cost, modular, and fully automated testbed capable of running multiple mesh protocols. It is also remotely accessible for any third party interested in performing network experiments with it. Apart from discussing the testbed architecture, we provide results obtained from a measurement-driven comparison of two protocols that fit within BLE operation: flooding- and connection-based. These results show that there is a trade-off between the two approaches and that the optimal mesh protocol to use is application-dependent, which should be taken into consideration when heading towards the next release of BLE mesh.