If you want it to work as a riving knife, you need to get it to around 4-5mm of the back of the blade, and be pretty well supported and rigid.
As you rip-cut wood (i.e. roughly along the grain), you release tensions in it from when it was a growing tree.
Think, for example, of the parts of the trunk where large branches join: mechanically there's an enormous bending force exerted on the fibres of the trunk - the tree grows to counteract them, to keep the branch out sideways. When it's felled, the force from gravity on the bough is removed. Those tensioned and compressed fibres will stay roughly in place when the trunk is planked, but they're like long springs embedded in the wood. And trees rarely grow perfectly straight, so there'll be twists and all sorts of odd geometry associated with the original shape
When you rip-cut, you sever some fibres but not others, and the wood moves as a consequence. The dangerous movement is anything that pushes it towards the rising teeth at the back of the saw blade. This might be the cut closing up, but it might also be any twisting action of either the stiock or the waste piece as they move past the blade.
It's why using a full length 'rip' fence on natural timber is extra dangerous - it doesn't help the straightness of the cut at all, but it does give the wood something to spring against, and the force can be released in only one direction - towards the dangerous back of the blade. So a short fence - stopping just after the front teeth is safer. It guides the stock up to the blade, but once the cut has been made (at the front of the saw blade) there's room for cut pieces to fall away, or for any wood movement that otherwise would pinch the wood into the back of the blade.
Cross-cutting doesn't create any of the same risks. If anything the wood will spring away from the blade, rather than towards it, and the geometry of what you're doing and where you are holding stuff means it's hard for the wood to be grabbed and launched
And that pinching is further mitigated by a riving knife - so the wood cannot get right on top of the teeth. Anti-kickback fingers (serrated pawls) don't do it, for the same reason that your splitter doesn't - they are too far back from the danger area, and anyway, rely largely on gravity or a spring, and on a dangerous kickback happening with a limited predicted range of motion (they only jam the wood against the saw table - what about if it's already above the table when they engage...???) .
Why a long fence, if it's dangerous? Answer: man-made boards. They don't have the odd tensions that natural wood does, and in this one case, the long fence may help keep the cut straight. In this context it isn't really ripping (despite the common use of the term), as the wood grain is in many directions, and not with any stored-up tension to be released. You can get kickback, but not of the same nature.
Hope that helps.
E.
