The elastic effect is basically if, for example, a player has one or two primary skills higher than the other(s), and when they try to train that/those skill(s) that is/are higher, it takes longer to pop because the primary skill(s) are so much more higher than the other primaries. The other way it works is if a player has a group of primary skills higher than one primary skill. Then the user trains that lower skill, and the skill that is lower starts to pop faster than the usual rate until that skill is close to the other primaries.
To answer your second question, I would focus on this allstar at first, because he has better skills to start with. He can be trained as an SF. Later on, I would focus on the other player. Lucky for you, the other player is an MVP, and can be trained longer.