In regards to drilling specifically, I personally prefer two methods. I like to drill individual games with 8 minutes on the clock, and sections obviously with 35 min on the clock. I like to incorporate the first method as a means of not allowing myself to compensate for a hard game with an easy game. There's always a game or two I finish in 4-5 minutes, so I usually have extra time later, but I don't want to come to rely on this, so I break them into 8 minute chunks, which leaves 45 seconds for bubbling and turning pages and such.
I don't believe in untimed practice for logic games. Maybe VERY early on when you have no understanding, but the games section is about efficiency and maintaining an idea of how things fit together. I fear taking the clock out of the equation leads people to practice inefficient techniques such as plugging in many answers to test them. Anyone could do games untimed, just chart all possible permutations, obviously this wouldn't fly in 8 minutes. If I'm still drilling a game with efficient techniques when the clock hits 0 I make educated guesses just as I would on the test, then I continue working through to solve for the correct answers past the clock, but I always get my first choices in within 8 minutes to see if I had enough of an understanding to at least guess the last few correctly.
My experience was a little different: I did a good bit of unlimited-time practice with games, and a good bit of brute-force permutation stuff in my practice, and both helped me tremendously in feeling like I really understood the games and their logical structure.