It is also worth considering that the takeoff demands the greatest performance of any phase. The airplane is at the heaviest weight and is expected to lift off at a predetermined point and produce a specified climb performance.
Boeing has never protected their tails. The premise is that they have used an airfoil that produces so much margin over stall throughout the flight envelope that it simply will not be vulnerable to icing. They certainly do the testing with three inch ice shapes, but the impression I have gleaned from numerous conversations with them is that their tails have a lot of capability in reserve (they don't like to get into details).
I believe that Airbus approaches this in the same way.
Contrast this with Douglas and the DC-9/MD-80. I believe that they started with the intention of not protecting the tail as well. Unfortunately, the stabilizer did not quite make the grade with ice shapes attached. In fact, the DC-9/MD-80 does have a demonstrated problem with ice contaminated tailplane stall. I am aware of two events ...I'm sure there are many more...in which a pitch over was experienced following the selection of landing flaps (the almost universal point of initiation for these types of events). I believe that the reason that no more serious events have occurred is because, unlike a turboprop, landing flap on the jet is usually selected prior to reaching the outer marker. This leaves plenty of room for recovery. In the Viscount/Jetstream/YS-11 events, final flaps were typically selected much closer to the ground.
In any event, Douglas had to bolt on ice protection for the stabilizer, which I believe is why there is not enough energy allocated to run both the wing and the tail at the same time. The required cycling of the tail ice protection prior to final approach is extremely important.
As far as the dynamics of an ICTS event, certainly the first manifestation is a notable reduction in longitudinal stability and perhaps a vibration in the elevator controls (as opposed to an airframe shudder from wing flow separation). The elevator snatch that Mad Scientist describes is next, but a complete tail stall is possible depending on the ice shape and tail angle of attack. In any event, a retraction of the flaps reduces the downwash angle and instantly reduces the tail angle of attack. Overpowering the elevator with brute force will re-camber the tail and restore flow attachment.
To date, this has not been a problem on any control surface that uses hydraulic power for flight controls (another reason Boeing is confident). Whereas a genuine elevator snatch will jerk the control column right out of your hand and smash the panel with it, the aerodynamic forces that generate this are easily overpowered by hydraulics.