How to define this larger frustrations? Well, there are a number of things that RPGs tend to model poorly: size differences between combatants, positional advantages, the importance of mobility in combat, 3-D combat (aerial, underwater, whatever), mounted combat, and a whole host of other things.
From where I sit, there have been four major schools of thought on how to deal with this. The first is the D&D method, which (from 3.0 on) squishes everything into two dimensions and pretends that height (and, these days, facing) doesn't exist. Tactical crunchiness is generated by using miniatures on a grid. 4e previews and the proliferation of powers that focus on moving opponents around on the grid make it look like this direction will, if anything, be emphasized in the future.
The D&D method is often fun, but there are a lot of things it just can't do without a lot of house-ruling. For instance, it can't really handle it when my halfling rogue with a +14 Climb check wants to climb up the back of the Frost Giant and slit its throat.
The second method is the Exalted method, which doesn't depend upon a tactical grid but instead gives you a certain amount of narrative power that can result in concrete bonuses to your action. In part, it is a solidification of a probably-common tendency in RPGs for the GM to give a bonus on actions that they deem cool. The Exalted method doesn't depend on rules crunchiness in combat (though, perhaps oddly, Exalted itself is pretty darn crunchy) to give PCs combat options. Instead, it depends on player creativity, GM willingness to allow players to take narrative control, and a shared vision of what's going on.
The third method attempts to account for EVERYTHING with hyper-specific rules-crunch (AD&D 1e? GURPS?), while the fourth just abstracts all complications away (OD&D?) and keeps combat simple.
I want to propose a fifth method. I'd be surprised if something like it hasn't been done before, but I can't think of an example. Basically, the idea is to do away with the need for a tactical grid and model positioning and such as a series of tactical advantages that fall into a couple of categories (such as "reach advantage;" "awareness advantage" - which would include gradients of surprise, readiness, and flanking; and "positional advantage").
So, for instance, my halfling rogue fighting the frost giant:
- The frost giant would have a major reach advantage
- The halfling would (coming from behind) have an awareness advantage (possibly negating the reach advantage)
- The halfling, after getting inside the giant's reach, would have a reach advantage. After climbing its back it would have a positional advantage.
Now I just have to make it into a coherent system...