No argument about the seriousness of kick back, but there can be practical problems with a short fence too.
First off there's a point nearing the end of the cut where there's not enough of the workpiece overlapping the short fence to be able to keep it guided parallel to the fence all that accurately.
This opens the possibility of the end of the short fence becoming a fulcrum around which the workpiece rotates if it's not kept tight and parallel to the fence - levering the last part of it out into the leading edge of the blade. Not a safety risk per se as its cutting down, but it definitely risks damaging the workpiece.
A full length fence avoids this possibility by fully supporting the workpiece. The far end can also be clamped down if needed to avoid deflection.
What is a risk (albeit probably minimal if the fence finishes towards the centre of the blade - but more so if the fence ends very close to the leading/cutting edge of the blade) is that when the trailing edge of the workpiece clears the fence it becomes very possible in the resulting space to rotate it with the push stick/whatever so that the far end pivots out into contact with the trailing edge of the saw blade - creating a kick back risk.
The riving knife should prevent this causing a problem, but it's probably another argument for allowing the workpiece far enough forward to properly engage the riving knife before it clears the fence.
A full length fence avoids creating the space for the workpiece to rotate, but of course creates the risk we hear of all the time by providing a surface against which the workpiece can be nipped into the rising edge of the blade. But equipped with a properly set up riving knife maybe the risk isn't that high?
I've worked both ways without incident, but don't have the experience to make a call on the relative risks/benefits. But it seems to me that both have their pitfalls, and that there are fine details of set up and practice that matter in reducing risk and getting good results in both situations.
Despite these shortcomings the realpolitik of the environment the safety industry in this part of the world has created prevents serious discussion - it seems to be that if you are an employer you keep your head down and do as the ideology specifies or risk all.
What's clear despite this though is that there are so many people using so many full length fence systems on so many saws in the US that they can't be anything like as dodgy as the safety brigade over here likes to portray them to be.
How could it be the case, given that it's the most litigation inclined society in the world, and given makers (who often will happily fit guarding so restrictive in some cases as to make a tool almost unusable so you take it off and so assume liability) bending over backwards to avoid claims???