in the latter and mixture of elements when heating to the extend the molecules join are considered and alloy or composite. Steel has less carbon than cast iron but also usually includes other elements to bring out additional properties associated with a stronger molecular bond. But the real issue here is the more carbon the greater potential for the intrusion of water and oxidation over time. When they were making canon balls they weren't all that concerned with quality control. Plus even if they were, in the explosion there are stress cracks in fragments that can only be seen under a microscope many time. These cracks allow in water that accelerates the corrosion. To stop the process you must remove the water by displacement (usually by some oil base material) or driving it off by heat which creates steam that forces the water out in the form of a gas. If there are significant stress cracks even though not necessarily visible and there is trapped water then yes when heating you can crack the frags you already have. Steam is one of the most dangerous forces and that's why boilers are high quality control devices. How much or how bad the cracking or flaking will be can't be estimated without spectrographic or x-ray diffraction testing. That's why long duration displacement methods are used for the most part. Think WD40 which means water displacement formula 40 out of the original 279 that were tried. Its more costly and time consuming and there's still the hybrid process of displacement with controlled heat and ventilation which can be dangerous not only because of fire potential but also dangerous gases released. But it does speed up the process considerably like a month over a year by comparison. Even then the displacement material has to be selected cautiously and strategically. Sometimes one process is used initially and this a different or third process before its over depending on the significance of the artifact.