Fault tolerance has become an essential capability in manipulator design methodologies as robotic manipulation systems are more frequently employed in hazardous environments and on geometrically complex or heavy duty industrial operations, where mechanical joint failures are likely to occur. This work focuses on the development of a redundant manipulator design methodology aimed at minimising the degradation in manipulator performance quality that results from k arbitrary joint failures. The relative weighted global isotropy index (RWGII) is developed for use as a manipulator design fitness metric. This metric takes into account the primary manipulation goal of maintaining kinematic dexterity, the secondary goals of collision avoidance and torque minimisation, and fault tolerance capability. The genetic algorithm search of an immense manipulator design space, conducted using the new fault-tolerant manipulator design fitness metric, yields redundant manipulator designs that effectively minimise fault susceptibility due to k joint failures while maintaining dexterous, redundancy-resolved motion on specific tasks.