There have been a handful of events involving 727 and 737 aircraft experiencing buffet after liftoff, and the 737 has had its history of pitch ups. There have been a couple of cases in which the 737 was unable to rotate, but inadequate deicing was not proven to be the case, only suggested. The compressor stall/FOD problem is huge; I have tracked quite a number of those, and the safety ramifications are clear when one considers the SAS accident.
A common occurrence in the data is to fail to deice the upper surfaces of the horizontal stabilizer, particularly on T-tailed designs such as the Dash 8 and ATR. This results in excessive pitch up tendencies, often using all of the trim as well as forward control displacement. Typically, the deice crew failed to adequately cover the tail, and the flight crew cannot see it under any circumstances.
It is true that all icing is not fatal; that is precisely the problem. Icing is highly variable in effect, and it is quite easy for a pilot to misinterpret his/her experience. Current research is being done, and has been done for several years now, on identifying critical parameters of ice shapes so that worst-case models can be built. We already know that a few thousandths of an inch of surface roughness can seriously outdo the large ice shape, so bolting two-by-fours to the wing probably doesn't tell us as much as we used to think. Some parameters under analysis are chord location of the horn, horn height and horn angle. It turns out that horn angle can have quite an effect, and I'd be willing to bet that regardless of your experience or eagle eyes, you can't see a few degrees difference between one ice horn and another. Neither can the icing tunnel engineers, which is why the measured data is what identifies the worst case parameters, as opposed to ten years experience working in the tunnel.
The upshot is that, like a lot of things in aviation, this is a matter of margins. Some guys end up using up all the margin and go off a cliff, which is reported in the papers the next day for all to see. A whole lot of guys operate with vastly reduced margins and never know it. Protecting the margins is what safety is all about, because you will never see the one coming that gets you.
Of course, to do that professionally, someone has to take the time to educate pilots on where the margins are and how they are constructed, so that he can make decisions specifically aimed at margin protection as opposed to single event avoidance. That is where the industry training is woefully inadequate. Lately we have seen ample evidence of this in landings which depart the end of the rather short runway, but the same principle is active in many, many icing events.