This is typical water hammer behavior. There is plenty of material out there to find on it, so I'll let you investigate further on that, and just answer your questions directly.
1. Sort of. In typical stress analysis, we're most concerned about the imbalanced force that causes flexure of the bends.
2. Input the loads into the elbows, opposite the direction of flow out of the elbows. If static, you should use a DLF of 2. Dynamic may become obligatory if:
a. This DLF overstresses your pipe but without it, it does not.
b. You have many more than 9 different forces to apply, and inputting them statically can result in loads falsely cancelling each other (although you could have multiple static models for remainder loads, too if >>9 forces)
c. If the individual forces interact in such a way that dynamically you will calculate stresses greater than individually, even with a DLF of 2.
I would suggest looking up Aft-impulse videos. For what it's worth, there's a video that covers an example of importing aft-impulse data to CAESAR that does a decent job at explaining the situation.