I just had to use Kevin's amazing icy water. ;)
nice one. To reduce the spikeyness, you could lower the amount of displacement in the ice layer or stretch the noise in x or y in the PF feeding the icelayer and/or try a different noise function. The displacement roughness and spike limit will also have an effect.
Experiment away!
seen like this it has potential for sea foam...must experiment
This might give you a fairly good idea for the look of ice formations.
One thing I've noticed is we keep trying to emulate that intersection areas, which only occur when ice sheets collide. It doesn't really make sense that they just occur, for example, on a lake. It does look cool.