You either don't work in IT or haven't worked there long enough.
I don't work in IT. School me.....
It's not unusual for systems not to be designed to scale well, particularly when the need to scale is hard to foresee. Sometimes it's just under-specced hardware, sometimes it's unfortunate architecture decisions, sometimes it's something as simple as neglecting to add an appropriate index to a table (I have seen that drop a query lasting over a minute to sub-second response). Then, not only is your system suddenly struggling under a load, it's so busy that it's hard to interrupt it to perform necessary upgrades. Sort of a catch-22.
What also doesn't help is that your system can be performing fine up to maybe 95% capacity then you cross a threshold and everything goes to crap. Queries start failing so people start hammering the system more trying to get their information instead of getting it in a nice steady flow. You start falling behind and things just go from bad to worse, you're spending your time catching up instead of servicing new queries and because of this, people are putting in more and more new queries. Your problem is generating problems.
Now, ideally, MtGox would have spent some time architecturing things so that things could have scaled easily. Then again, ideally Cyprus wouldn't have set things up so that they had to dip into peoples bank accounts so...