That I'm aware, intersection tests are done using the geometry prior to having the individual pixels. You don't need to worry about rasterization until near the end when you're trying to determine what color a pixel should have based on what you've discovered. And really, you're talking about a sort of antialiasing. In ray tracing, that basically amounts to jostling the camera by very small amounts and averaging the pixel colors so that you end up with pixels that better reflect what you're raytracing.
Strictly speaking, you don't need this in order to have a ray tracer. You should focus on getting an image first. Once you have that, then you can apply antialiasing on the image data obtained.