Brett O'Callaghan writes:

As a "software person" (;-), I will say I'm "surprised" this issue
didn't come up in testing.

As in faster-than-real-time simulation of the whole mission --
many months in space and many months on Mars, presumably.

But from what little they've said of their problem, I will say I'm
surprised this issue didn't come up in analysis and/or that they
didn't keep track of the RAM usage as they caused flash files to be
created. I'd think they'd know how much RAM each flash file needs
(maybe depending upon file size), etc... They probably trusted the
sorta-off-the-shelf software too much without understanding it enough.

But maybe they did all that and it still went bad. I infer from the
brief briefings that they still don't understand why the flash files
needed as much RAM as it seemed to need. The problem might have only
been seen or tested for in some very special and poorly-understood and
hard-to-simulate conditions. Complex software is like that and it's
practically impossible to know how much money and time to spend on
understanding and testing the software, even if one has any choice.