The problem of course, is that the travel time for the command to get from Mars to Earth for the controllers to make a decision is 4.3 minutes at the shortest distance, and as much as 21 minutes at the longest. So with a round trip of ~10 minutes best case, the mission is over before the command can be executed.
This brings up a point that I often make in my classes. When you're designing a spacecraft or planning a mission, it is
always necessary to ask, "what's the worst that can happen?" A lot of times the answer is "not much," but in some cases -- autonomous landing on a planet, for instance -- the answer is, "loss of mission."
For such cases, it is necessary not only to carefully
design, but also to carefully and comprehensively
test that portion of the mission to ensure that the risks of "loss of mission" are understood and, to the extent possible, mitigated or eliminated. This imposes costs, obviously, but the question is: does loss of mission outweigh the cost savings? (In some cases, cost may actually win the argument.)
There are numerous examples of what happens when you don't ask the question: The Mars Polar Lander failure could have been identified and mitigated through testing and software dynamic simulation, but that was deemed too expensive. The one test that would have identified the "wrong prescription" for the Hubble telescope, likewise. The recent SpaceX explosion destroyed the payload, for no good reason other than considerations of cost and a couple of days of schedule. Perhaps this lander failure had the same root cause.....
But it seems that they maybe either didn't ask the question (and therefore identify the problem in test), or their cost/risk decision was ill-founded.
Human on board is the best fix for programming unknowns.
Absolutely. The only problem being that it's really expensive to keep humans alive....