[wiredfool]

When discussing the 2 rovers with a 90% success rate, the author comes to the conclusion: <blockquote>The right answer is to go for two rovers, because if you do it that way, you will have a 99 percent probability of succeeding with at least one of the vehicles</blockquote>

Which is not exactly right. It's correct if you're looking at random, uncorrelated factors. However, two rovers from the same program are not going to be uncorrelated. If one rover is hit by a software blunder, it's likely the other one will have the same problem. (e.g. using mks instead of english units in the flight computer, using a 16 bit counter that overflows to name two)

[Dove]

If one rover is hit by a software blunder, it's likely the other one will have the same problem.

Actually, this is something they plan for. It is not unheard of to have the software for two different devices developed by two different teams who are not allowed to talk to each other, precisely to avoid this scenario.

In fact, there are Common Cause Analysis people whose whole job is to think of this sort of thing and recommend ways to avert it.

[Symmetry]

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.

[rflrob]

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.