Ok, you are correct that a tiny piece of the fluid has one velocity, but what about the tiny piece of fluid next to that one? Remember, before we take the limit as area goes to zero, we have a tiny, but finite, area. However, once we take the limit, we are dealing with one point in the fluid. So, there is a velocity distribution over a larger area, then as the area is made small, the variations look linear, then if the area is made arbitrarily small, zero crossings can be eliminated.