Or they could not launch the two rovers at the same time and use the failures of the first rover to make the second work, in which case the errors become anti-correlated.
The problem with that approach is that given the orbital dynamics of the Sun-Earth-Mars system, there's a relatively brief window every couple of years when it makes sense to launch a mission. If there's a crippling error in the first launch, it's somewhat unlikely that it can be identified and fixed before the second one hits Mars orbit.