The surge "hits" the downstream surface at the "end" of each straight run of pipe. These are individual event that occur one after the other. They may be linked dynamically or the response to each event may be isolated. The first "end" is the valve face and the event duration (the imbalance) is the length of the straight run divided by the speed of sound in the fluid, when the pressure wave front passed around the upstream elbow, the first event ends and the second event begins. This second event lasts until the pressure wave front enters the next change in direction. and so on.
Don't think of "two directions" at elbows. The two directions are two separate events. We are working with delta P here not change in momentum. (The loads associated with a change of momentum will have the "two directions" working almost simultaneously.)