You're right that the 2D representation doesn't work very well. The thing is, the 3D picture you have in your head also doesn't work very well; in order to make it work, you need four dimensions. Remember that general relativity is about "spacetime", not just space. Nothing is ever actually stationary; everything is always traveling at C. If it's stationary in space, it's traveling at C in the time direction. As it moves in space, it diverts some fraction of its motion in time into space.
Now, given that picture, we can reduce it back to two dimensions to make it possible to visualize. Imagine an infinite plane with two dots moving on parallel tracks at identical speeds. Overlay a grid over the plane to make the warping easier to see. Now, rather than distorting the plane in a third dimension like the typical rubber-sheet analogy, just "pinch" the grid lines inward towards each of the objects.
If there's only one object, the effect is totally symmetrical and doesn't affect its trajectory at all. If there are two objects, though, each one starts to run into slightly skewed grid lines caused by the other one. In an attempt to follow a straight-line path, they end up curving towards each other, and the effect gets stronger the closer they are to each other (because the grid lines get more and more bent). Eventually, they collide and electromagnetic forces overwhelm the gravitational forces that are trying to make them occupy the same space.
To say that again, it's hard to imagine with stationary objects, but there's no such thing as a stationary object in spacetime. With moving objects, you can imagine how they'd warp towards each other; if all their initial motion is in the time dimension, that warpage basically rotates part of their motion into the spatial dimensions. That's also why gravitational fields cause time dilation, because they rotate the "stationary" direction away from the time axis.