Some people love watching people get hurt, they may find it humorous. Some people don't really like it but can watch it. Some, however, find it unbearable to watch. Why is this? Are the latter of the three categories of people listed have more mirror neurons than the average, or are they more dynamic, perhaps? Do the first type of people listed have less mirror neurons than most, similar to those with SM, who don't perceive fear, or are their mirror neurons less active? If we all have these mirror neurons hard-wired into us, as humans, why are the impacts of our mirror neurons so variant among differing individuals of the human race?
An inability to watch others in pain can be described as an empathetic reaction, which mirror neurons aren't necessarily responsible for. Mirror neurons are hypothesized to facilitate 'action understanding'. In other words, they are able to recognize the intentions of others' actions, thereby allowing them to anticipate their future actions. What may be happening is that stronger activation in the mirror system for emotions may be a contributing factor to empathy and moral thought.
The VMPFC (ventromedial prefrontal cortex) is sometimes considered the 'moral network' of the brain, and mediates the automatic reactions which occur when one must face moral decisions . Patients with emotion-related damage in the VMPFC have been shown to express more utilitarian judgments than average, which judges not based on the perceived harm of the individual but rather the negative objective impact on the world at large, and thus may be less readily affected emotionally by suffering, particularly at the individual level.  Other studies have shown that rational processes mediated by the dorsolateral prefrontal cortex (DLPFC) can override the moral network  and lead to utilitarian judgments.
 Marazziti, Donatella, et al. "The neurobiology of moral sense: facts or hypotheses?." Annals of general psychiatry 12.6 (2013).
 Moll J, de Oliveira-Souza R: "Moral judgments, emotions and the utilitarian brain." Trends Cogn Sci 2007, 11(8):319-321