About metal fatigue
When you repeatedly apply force to a piece of metal, in general, it will eventually fail. The higher the force, the fewer cycles to failure, and the lower the force, the more cycles until failure. However, when the force falls below a certain threshold, called the endurance limit, you can repeat the application of force forever, with no failure.
The actual measure for the endurance limit is in stress, not force. Stress is force per square inch. This means that if something is stressed above its endurance limit, you can always make it last longer just by making your metal thicker. You may even make it strong enough that you drive the stress below the endurance limit, and then it will never fail (under normal usage).
For aluminum though, the endurance limit is always asumed to be zero (if it is non-zero, it is too small to be useful). You can never get below the endurance limit, so eventually, every aluminum bike frame will fail. The reason consumers accept this is that (hopefully) aluminum frames are made with the metal thick enough that the fatigue failure doesn't occur until peak cycles get up into the hundreds of millions, or higher.
For steel, however, the endurance limit is comfortably distant from zero. Practical steel supports can be designed that should never fail during normal operation. The endurance limit for steel typically comes into play on the order of one million cycles; in other words if something has survived a million cycles of stress, it should never fail (due to that level of stress). For a bicycle at a rather slow cadence of 60 RPMs, this is less than 280 hours of riding.
There are certainly many riders capable of accumulating this many hours of out-of-the-saddle riding time in a single year, and a few who could get in a million cycles (out of saddle) in just a couple of months. And most of these riders have frames which have lasted them much longer than this. This is simply not the time scale in which frames should be failing. From this I conclude that steel-framed bikes are generally designed with all normal usage falling below the endurance limit.
One of the things this means is that if you do have a steel frame failure, it is due to one of three things: You exceeded normal usage (e.g. crashed), or the frame had a manufacturing defect, or the frame was poorly designed. Note that crashing may not cause immediate failure, but it can lead to a failure later on. For example, a crack in a frame will redistribute stress to the areas around the crack. These areas will see significantly more stress than they saw before there was a crack, and hence may be prone to an endurance limit failure.