You might want to take a look at Lonnie Athens (Wiki - brief and incomplete outline of his theory of violence) - focused first on his research of violent criminals in American prisons. His general conclusion is that, to understand violence, experience is a more important factor than logic, ideology or genetics.

Athens' theory has been considered by Richard Rhodes - generally in Why They Kill: The Discoveries of a Maverick Criminologist (a brief review of Rhodes' book - but explaining Athens' four stages in a nutshell); and specifically with respect to the SS Einsatzgruppen (in one chapter of Masters of Death).

In both books, Rhodes concludes:

"He now firmly resolves to attack people physically with the serious intention of gravely harming or even killing them for the slightest or no provocation whatsoever. . . . He has suddenly been emboldened and made venomous at the same time. . . . The subject is ready to attack people physically with the serious intention of gravely harming or killing them with minimal or less than minimal provocation on their part." Says Rhodes, "that is, he is ready to become an ultraviolent criminal."
That is a valid comment re: genocides, serial killers, multiple killers and the "mind of the murderer" in general. As to genocides and Athens' theory, see Dimensions of Genocide: The Circumplex Model Meets Violentization Theory, by Mark A. Winton (2008, case study of Rwanda).

The bottom line is that Athens' theory generally holds up (although as Winton points out, it is non-exclusive and ties in with other approaches - including some of Grossman[*]) with respect to "bad guys".

However, if it is truly a "general theory", it should also hold up for "good guys" - with adjustments in terminology. Athens gets into that in Violent Encounters: Violent Engagements, Skirmishes, and Tiffs (Journal of Contemporary Ethnography, 2005), in his Conclusion:

CONCLUSION

The interaction between perpetrators and victims when violent crimes are either attempted or completed can be best understood if it is seen as arising during social acts—activities that require the voluntary or involuntary participation of at least two parties for their completion. With the obvious exception of suicide, all violent crimes constitute social acts because there must be at least two parties for them to be committed—a perpetrator and a victim. Lawful violent acts constitute social acts just as much as unlawful violent acts do. In the cases of excused and justifiable homicides, as well as excused and justifiable batteries, there must also be at least two parties—a perpetrator and a victim. The same is also true in intergroup violent criminal and noncriminal violent action, except that in this case, the victims and perpetrators are collectivities rather than individuals. In collective social acts, it is groups rather than individuals who perform the separate roles, communicate through their spokespersons, assume each others’ attitudes, and try to work out a congruent social object or plan of action for carrying out the larger social act in which they are the acting agents (Blumer 1966, 540; 1969, 52, 55-56; 1981, 148).

As in the case of individual social acts, there are two kinds of collective social acts: cooperative and conflictive. Unlike in cooperative social acts, in conflictive ones, the acting agents, no matter whether they are individuals or groups, cannot form a congruent social object or plan of action because they cannot agree on who should perform the superordinate and subordinate roles in carrying out the social act. Unsurprisingly, violent encounters do not arise during individual or collective cooperative social acts but instead during conflictive ones. It may be speculated that the violent encounters that emerge during either individual or collective conflictive social acts fall into the same three basic subtypes that differ in terms of the number of the five stages of a violence encounter that are completed: (1) role claiming, (2) role rejection, (3) role sparring, (4) role enforcement, and (5) role determination. During a violence engagement, all five stages must be completed; during violent skirmishes, only the first four of these stages must be completed; and during violent tiffs, only the first three must be completed. Thus, despite the differences in legal status between lawful and unlawful violence and between individual and collective acting units, the grounded theory of violent criminal social acts that individuals perpetrate described here could be potentially applied to violent social acts that are both lawful and unlawful and that both groups and individuals perpetrate and, thereby, to all violent social action.

Before this extrapolation can be safely made, however, appropriate amendments would undoubtedly have to be made to the theory. Any general theory of violent social actswould have to take into account the added complexity that an increase in scale in the social act’s acting units would introduce into the proposed explanation (Blumer 1981, 148-149). Undoubtedly, the nature and size of the groups involved in a dominance encounter could significantly affect the actual social practices at work during the different stages.

As Blumer (1959, 129-30) pointedly observes, large and small groups must utilize different social mechanisms to perform their roles or “mobilize for action” in social acts:

“A. . . reflection of the collective factor in the case of large groups is the organization on which they must rely when mobilizing for action. A small group uses confined, simple and direct machinery. Corporate action in a large group requires the articulation of more units which are also likely to be more diverse, more removed from each other, and related through bridging links. . . . The mobilization of this extended, diversified, and indirectly connected organization requires forms of leadership, coordination, and control which again differ from those in small groups.”
Of course, a general theory of violent social action also could not ignore the state’s approval or disapproval of the use of violence. Obviously, this is a factor that could also significantly affect both individual and collective acting units’ performance of their roles in violent social acts. Thus, future research would be needed to determine the exact nature of the amendments that would need to be made in each of the stages through which violent engagements, skirmishes, and tiffs pass to accommodate all violent social acts rather than only the criminal ones that individuals commit.
--------------------------
[*] I've read both of Grossman's books (On Combat and On Killing) - many parts of them more than once, since they obviously relate to actions that are in some cases "war crimes" and in other cases not. I believe Athens' theory is consistent with much of Gerossman's factual material - and also ties in with current training theories and practices for military and police.

Regards

Mike