It is important in my idea that it is a comet - that is large chunk of not connected ice, not a hard rock. As we know from examples of Tunguska event 1908 and Chelabinsk event this year, such bodieas tend to desintegrate on high altitude in large explosion because they quickly fall apart. Chelabinsk flash video. https://www.youtube.com/watch?v=OPSzpnHHwos
Estmation of the energy is based on the speed of imact, that is 600 000 meters per second (second cosmic speed on the Sun surface), mass of the object - that is 10++18 kg (based on water density and size of a cube with 100 km rib) and formula for kinetic energy, that gives us enegry of impact 3.6x10xx29 J.
Energy output the Sun is 10x26 J per second. So, total energy of impact would be 3600 times more than Sun's output. Not all energy will go in radiation so 1000 times seems to be good estimate.
In fact I started from the question «What is the size of the body, which could cause harm to the Earth if it fall on Sun?" And find that 1 km will not be even visable, but 100 km is dangerous.
Now I need estimation of the frequency of such impacts.
First correction, the Sun's luminosity is ~3.827E+26 W, so the falling ice-cube would have a kinetic energy of ~1,000 seconds of solar output. However an object falling into the Sun - such as a 100 km ice-cube - would only release its kinetic energy in such a burst if it was brought to a sudden halt. At 600 km/s the object is a solid surface moving through a fairly diffuse gas - the outer layers of the Sun are thin, hot plasma. A good estimate of the braking effect would be Newtonian Flat Plate drag - i.e. the stuff of the Sun immediately in front of the o...
Bolonkin & Friedlander (2013) argues that it might be possible for "a dying dictator" to blow up the Sun, and thus destroy all life on Earth:
Warning: the paper is published in an obscure journal by publisher #206 on Beall’s List of Predatory Publishers 2013, and I was unable to find confirmation of the authors' claimed credentials from any reputable sources with 5 minutes of Googling. It also has two spelling errors in the abstract. (It has no citations on Google scholar, but I wouldn't expect it to have any since it was only released in July 2013.)
I haven't read the paper, and I'd love to see someone fluent in astrophysics comment on its contents.
My guess is that this is not a risk at all or, as with proposed high-energy physics disasters, the risk is extremely low-probability but physically conceivable (though perhaps not by methods imagined by Bolonkin & Friedlander).