I agree with fermiboy. Several users from the past few years have commented that the types of problems where you can get the right answer just by knowing what units arise when you divide a tesla by a newton seem to be a thing of the past. I didn't find any such problems on my test. A lot of my problems were the kind where all the answers were the same except for the coefficient,
1/4 kg m / C* s
1/2 kg m / C* s
1 kg m / C* s
2 kg m / C* s
4 kg m / C* s
or they asked for the ratio of two quantities, and there were lots of purely conceptual questions with sentence answers.
Perhaps the dimensional analysis strategy was useful on the early version of the released GREs (8677), and people discovered the method after that exam was released. But if I had to guess, I'd say once ETS realized people caught onto it and could answer problems without solving them, they wanted to put an end to it.
However, the method of limiting cases remains very useful! I think at least 3 times on my test I answered a problem that seemed too cumbersome or difficult just by letting d-> inf, or n(index) = 1... only one answer worked, and if two answers work, try a second limiitng case, if possible, or just guess. It's almost as if they intended us to solve them that way. That woudln't surprise me, since applying limiting cases requires some physical intuitiion and is a skill that a strong student/physicist should have.
Also, knowing when an answer should increase or decrease with a given variable is important too. A classic example is the problem (on one of the practice tests) of the marble rolling along a sinusoidal track like Asin(kx) and it asks you to find the speed at which the marble will fly off the track in terms of the constants. Think about it, as k increases the wave gets tighter so v should decrease, as k -> 0, the track becomes flat and v -> inf. As A increses, v should decrease too etc...
Anyway, the lesson of the story is that nobody planning to take the test this fall should go in there expecting to find too many problems that can be solved by magic tricks without thinking. Also, note that even computing units can be cumbersome, especially when working with constants such as the permittivity of free space (
yikes) which is in all the E&M questions. And having to divide nested fractions like (kg m/s C )/(kg m/s^2 A) can lead to a lot of careless errors.