The scenarios of interest for estimating bottom and sub-bottom physical properties now encompass both deep and shallow, or very shallow, coastal waters, for the deployment of renewable energy platforms (e.g., wind farms and wave/tidal energy plants). This new paradigm, together with the continuous requirement for reducing survey time (and cost), spinned out the concept of a distributed and reconfigurable seismic survey system composed of a fleet of autonomous underwater vehicles (AUV) carrying acoustic sensing arrays. Such system poses a number of technical as well as scientific challenges, among which that of sensor positioning for optimal bottom inversion performance in a given scenario. The present work addresses this issue through the eye of the sensor configuration that maximizes diversity and proposes sampling incoherence bounds for 1, 2, and 3D array systems. Random sampling is a concept that favors diversity and allows for the usage of low-complexity inversion schemes such as those based on compressed sensing. Simulations on realistic environments illustrate the proposed concept. [This work is part of project WiMUST—Widely Scalable Mobile Underwater Sonar Technology funded under program H2020 of the European Union.]