I simply find it difficult to believe that the Mt Gox system doesn't scale well enough to handle a high volume influx of trades;
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 to be not 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...
I am not a programmer but have been around it enough to understand the basics. My question was really 'why hasn't Mt Gox done anything about this?', but I do realize that it can be hard to make time for upgrades considering the increased volume they've experienced. I have seen very poorly structured databases handle extremely large volumes of data (a GSM/CDMA SCADA provider with tens of thousands of units in the field communicating mountains of data to a DB created by an electrical engineer with little programming/architecture experience). Of course they probably had better servers and I/O support. They hired a DB architect to clean it up after several system wide outages.
I'm simply shocked that this has not yet been addressed at Mt. Gox (or perhaps it's in-process). I guess it's much easier said than done; they weren't ready for this level of volume.....still, this doesn't seem like their data would be a very complex.