Orbital photographic and remote sensing surveys of the Moon and Mars show evidence of lava tube formation. Lava tube caves, including collapsed tubes with skylights, are of enormous importance to geology and astrobiology because they (1) allow direct examination of bedrock, potentially including materials brought up from depths that are otherwise inaccessible from the surface; (2) provide good protection from solar proton event radiations; and (3) by analogy to Earth, might provide access to a rich biosphere hidden from the surface, specifically adapted to life in that extreme environment. While previous mission paradigms have prevented exploration of lava tube caves on Mars, new robotic exploration technologies and paradigms can now make this possible. A potential point of failure for current missions arises from the use of a single agent (e.g., single rover) to conduct the entire mission. Such an 'all-in-one' rover must carry onboard a large and cumbersome array of instruments, sensors, computers, and communications equipment. Moreover, this type of rover must be remotely commanded and kept away from hazardous environments for fear of damage or overturn, thus incurring the loss of the entire mission. Such inherent deficiencies hold back the scope of the science mission for which the rover was originally deployed. However, these restrictions could be mitigated using an alternative multi-rover architecture for autonomous robotic space exploration. This system, termed 'Tier-Scalable Reconnaissance', would combine mobile robotic surface probes coupled with a base rover, in situ operational autonomy, and multi-tiered science reconnaissance to permit robotic traversal of risky, science-rich environments in a way that the loss of one or more robotic agents need not jeopardize the entire mission. As a prelude to actual planetary exploration, there are ready analog environments on Earth where such a multi-rover framework could be pre-deployed, tested, and evaluated.