Fatigue failure of steel spokes is quite common, as is the rim cracking at spoke nipples. There's people over at yacf who'll bore you to tears on the subject...
I would argue that most broken spokes and cracked rim eyelets are not true fatigue failures. And it's also a non-issue in the sense that it doesn't put a rider in hospital, or even stop them from riding home on the other 30-odd intact sets of spokes/nipples, unlike a fork failure which you'll be lucky to walk away from.
...An aluminium fork will eventually fail through fatigue (corrosion is more likely to do for steel). This isn't a laboratory property - it's a fundamental property of the material...I would expect most forks to last for many decades - unless someone got something wrong, of course...
I do not hear much about fatigue failures in metal forks, aluminium or steel, which is to say I think it is very rare. People who buy up old frames for "conversion" don't worry about how much life is left in the forks. I can only surmise that the designed fatigue life (i.e. thickness of metal and stiffness of the structure) of metal forks is sufficiently long for it to be a non-issue. As you say, the fatigue life of metal forks is probably in the decades, and I am not confident CF components will last decades - they haven't been around that long on bikes, and there are long-term effects (e.g. UV exposure) that are slow but, like metal fatigue, cumulative.
...There are no shortage of real world catastrophic failures through fatigue - that's why engineers spend so much time worrying about it. How long depends on the design and quality of manufacture...But new designs come out in metal as well as carbon - how can you be sure the design is right there?
...
Engineers have "worrying" about metal fatigue for a very long time and it's now very well understood and thus easily avoided. Of course a new metal fork design can be dodgy, but that is rather unlikely, IMO, because there are very few metal fork manufacturers around (rather like the situation with groupset manufacturers) and they are rightfully very conservative in their design because they're big enough to be worth suing. CF, at the other extreme, positively encourages small-scale, piecework manufacture and the result is widely varying standards of design and fabrication. The shocker here is that a big company, Trek, screwed up (and their abysmal response to it) - what of the numerous smaller concerns?
...Fatigue is progressive. It takes months or years. With luck, you ought to be able spot a crack before brittle fracture.
That's another thing that makes CF problematic - delamination is often buried, where you can't spot it with a visual check, and the failure is sudden and catastrophic. With metal fatigue, you have a sporting chance of seeing it, and, especially with steel, bends before it breaks.
At some point in the future when all the quirks are known and ironed out, I think CF will be the material of choice, but I personally don't fancy being a guinea pig for CF "prototyping" failures.