=> in the liquid, individual water molecules are organized into transient ring structures, in which rotation is "hindered": that is, they can't rotate without banging into a neighboring molecule. Thus, there are no resonant transitions in the microwave region. Absorption takes place when the random motions of neighboring molecules allow a given molecule to follow the applied field for a short distance, after which it bangs into a neighbor, converting the motion into heat. This process involves all sorts of orientations of the molecules, and doesn't have any specific characteristic frequency but instead a wide range of frequencies: in particular, 2450 MHz plays no role, and 2300 or 2600 MHz would work just as well (but in the US the FCC would be after you). The use of 2450 MHz is a historical artifact having to do with frequencies licensed for industrial use, not related to any specific property of molecular or liquid water.
The absorption vs. frequency graph is in the web site I provided previously; it shows that as the water heats up, the peak in absorption increases in frequency, from about 15 GHz in cold water up to close to 100 GHz near the boiling point. The microwave oven frequency is chosen to be below the peak absorption point, so that as the surface of the cooked object heats up, absorption will decrease, allowing the microwave radiation to penetrate into the center of the object and heat it up and thus encouraging more uniform cooking.