I can't say that none of my own IOTs have crashed and burned: MIOT was never going to get off the ground, I never planed for it to (However it did get some of the best IOT advertising I've ever seen, including blackmail and stuff). I know why it failed and I'm aware of what not to do next time.
But it seems that saying that IOT games inevitably fail is an idea made from observations of IOT by people like Christos, Civoasis, Farshight and METY (no offence meant to those people). I'm sure if you were to look for a correlation between number of IOT started by a user over an amount of time and length of that IOT you would see a link.
However, if you look over the less successful IOT there are other ones that have worked: I&B, SonRisk, Sons of Mars and certainly the chronological IOTs.
I think that placing all of IOT culture under a banner of constant failure is wrong. All the games may have a mean turn length of 3 turns but there are ones that last for much monger and those are the defining ones - the true IOTs, the ones that make IOT what it is and defines it as a Sub-Culture of Forum Games rather than just a part of it.