AFAIK, in 12.x the main train loading design options were a (a) 6 logistic chests with fast inserters into the wagon, or (b) an n-to-6 belt splitter with fast inserters into steel chests into the wagon. In 13.x, logistic bots became more expensive while the stack inserters are a much improved way to load from belts. So, it seems that it is might be viable to revisit the inserter train loading design.
It seems to me that a much simpler design of a 1-to-4 double splitter to 4 stack inserters buffered with 4 steel chests suffices for a pretty high throughput:

What design do you use nowadays?
(related question: Do we have numbers on how many stack inserters can fully saturate/empty green/red/blue belts, and/or the number of items per second stack inserters can move to/from containers and belts? )