robby, maybe you should be NASA’s team leader for the next Mars mission so they can avoid these mistakes.
So, given the limited testing time available to the team, what would you suggest? Speaking as a design engineer, I believe a dead clever idea would be to design the robot to be reconfigurable in some manner, capable of dynamic updating during deployment, enabling the team to fix a large array of problems as they arise. Yes, that ought to do it.
But what’s that you say? They did? Well, goodness me, what clever chaps. And they seem to be successfully fixing things? Golly, it’s almost like these people are rocket scientists.
Er, yes - and? This seems like an eminently reasonable conclusion to me. Do you think they’re just going to wait until it fucks up, or do you think that they’re going to take pre-emptive steps to avoid problems?
I should remind you that the files believed to have overloaded the robot were accumulated during the flight between Earth and Mars. This is not a matter of the difference between a 9 and 18 day test; to have recreated these conditions the engineers would have had to have wrapped up the craft, stuck it in a cupboard for several months and waited. Not, I think you will agree, a sensible use of limited testing time and resources. You, obviously, have the grand benefit of hindsight, but please try to accept that not all eventualities can be foreseen, nor can every scenario be tested, no matter how much JPL may be lacking in robby-esque genius.
Well I’m just pointing out what seems obvious, the reporter didn’t quote anyone. I don’t doubt that she actually said 9 days of testing. You read more into the statement than I did.
Besides, how many EE grads do you know that move in journalistic circles? 
The fact is unless you can get the test cases run by JPL, or whoever did the testing, we don’t know. 9 days of testing on the memory management system dealing with flash seems reasonable. That is unless for some odd reason you think that they should’ve taken the rover system and run it through a complete simulated Mars mission on Earth. That would be about 200 days. Or maybe build a dummy rover that could’ve been built to be stress tested. That would’ve added another 200-300 million though.
As an aside, if you actually do get a copy of the test plan I’d love to read them.
Well personally I’m shocked that they did no near destructive testing on one of their only two platforms.
From what I can gather from 3rd hand reports, the problem isn’t the amount of data in flash, but the number of files, which would seem to indicate that whoever was in charge the file allocation table picked the wrong static size for an array (or something).
It was my understanding that the rovers launched w/o all required software on board (thought I’ve read so much stuff about this project in so many places I can’t cite this). This would make things more understandable, as I’ve yet to see a programming project that was 100% finished before the deadline. 
It’s not like this never happens for anyone else… Pathfinder had the same OS, and a problemwith priority inversion that wasn’t a major problem because there were always backup methods to contact the landers (as is the case with Spirit and Opportunity). Of course, Mars Polar Lander didn’t survive so well, but I guess this just means the major debugging focus should be on the EDS bit of the mission. You’re always going to run into software problems in large scale projects with one-off code… the exceptions being cases like the shuttle, but there the development costs and time were way more than what the current rover mission has to work with.