This document discusses testing uncertainties in cyber-physical systems (CPS) that span Internet of Things (IoT) and cloud infrastructures. It proposes combining model-driven engineering and elastic execution techniques to dynamically provision both the CPS under test and testing utilities across various IoT and cloud infrastructures. Specifically, it suggests using software-defined IoT units and cloud-based elastic services that can be composed, controlled via APIs, and provisioned elastically to enable testing CPS configurations and behaviors across heterogeneous environments.