The first is for fun. See the miniature lasso that he used? It is easy to tie one out of a long blade of grass and catch lizards that way. The trick is that they've been adapted to ignore grass as a non-threat. So when they're about to be captured by a piece of grass, they literally can't see the grass. You have to remain far enough away, but you can bonk the lizard on the nose, pull your grass back, readjust and try, try again until you get it. It's a great trick to show kids!
The second is that no mechanism is needed to revert a temporarily selected trait. Francis Galton was the first to document regression to the mean in inheritance. The child of two tall parents tends to be not as tall. The child of two short parents tend to be not as short.
The reason this happens is that the recessives that influence the one parent's trait are often different from the recessives for the other parent. Therefore the child winds up expressing fewer recessives than either parent.
What this means is that if you select for a given complex trait in a given generation, those who survive will have that trait. But unless you also select for it in the next generation, the population will mostly revert back on its own. It is only regular selection, sustained over time, which causes the selected for trait to breed true.