1) You are not missing anything. All the functionality related to variables mapping is basically within those two methods.
2) It should not lead to endless loop since setting a variable first checks if the new value is != to the existing value.
3) The consideration was for subgraph exposed variables to work similar to how code function parameters work, meaning that they are only given once when the function is called. It just made more sense to me to write-in only when the subgraphs starts, but I suppose that an option needs to be added to let us select in inspector a “once” or a “synchronous” write-in (and probably read-out as well).
