I gather from that Symbian had become unmaintainable. The usual metric for this is the number of bugs keep growing, despite strident efforts to fix them. As a long time Symbian user I can testify that was true. As a software engineer I can only guess what the reasons for that might be. I also gather from the link the great white hope was to replace Symbian with Linux (be that Maemo / MeeGo or whatever). So I was taking from that they were betting the companies smart phone future what was Maemo.
When the N900 was released it looked to me like that were going to pull that off. The incessant loud GUI arguments we seem to have about colours, button sizes and associated fluff were obviously still ongoing, but from a software engineering point of view the underlying libraries were working like a charm. On hardware that was more limited than the iPhone, the thing was crisp, the animations were smooth and all the normal functionality you expect to find on a phone was there.
Then two really odd things happened. If you are betting your company on an almost working platform, do you then: (a) move from a .deb distribution to a .rpm one, header by another company and (b) move from gtk to qt? Of those two, the gtk --> qt thing looks by far the worst, as all their gui has to be laboriously ported. At least there seems to be very good software engineering justifications for making that switch, even if the timing looked to be spectacularly poor. As it happens Nokia trolls swear they were ready for release by the due date - Q4 2010. Looking at the QT release dates - 4.7 released September 2010, I can't help but wonder if that view isn't distorted by a tiny amount of "it's my baby" bias, but it's hard to tell.
I can think of no software engineering reason for the Maemo --> MeeGo move, which was essentially a move from .deb to .rpm. But they are for all intents and purposes equivalent, so technically it wasn't a big move. If I had a Debian based system which I was working feverishly on to get the next release out, and was told in the middle of it I had to move it to RH/Suse/Whatever it would be done in a few man months, assuming I wasn't fired in the mean time for hurling abuse at upper management. The move isn't rocket science. If you take a scorched earth approach, you just use whatever packages are the same, port the rest, and package a appropriately configured kernel. There must be 1000's of engineers out there who could do this in their sleep.
Yet the Nokia trolls were quite emphatic on the Maemo --> MeeGo transition being the killer. Maybe the it wasn't the engineering, maybe trying to navigate the social aspect of going to a new distribution run by another company was like trying to swim in treacle. But regardless, technically it makes no sense. As bronson said was started 3 years before the iPhone, it was mature, and the technically the Maemo --> MeeGo should have been a minor bump in the road.
Someone who was there should write a "The Soul of a New Machine" book about it, so the rest of us can understand what happened.