Syria’s ongoing civil war epitomizes a challenge third parties – whether aid NGOs, international organizations like the Arab League, and indeed foreign governments – face in policy-making with respect to conflict: many civil wars are fought not between *the* government and *the* rebels but rather diverse armed actors. Uncertainty surrounding the trajectory of the Arab uprisings even in countries where regimes fell, due to factionalism and the inability of governments to secure a monopoly over the use of force, suggests the continuing relevance of this challenge. Multiple militias, paramilitary groups, and autonomous segments of the security forces leave an imprint on internal warfare. Observers sometimes refer to “kaleidoscopes”, “dizzying” numbers, or “countless” warring groups. Outright fighting between putative allies is particularly puzzling. What determines whether allied armed groups can successfully cooperate, using this to their advantage against a common enemy? Why are some wars prone to seemingly self-destructive fighting between purported allies, to the detriment of their shared interests and their civilian constituencies? This paper develops a concept of violent conflict within alliances, an understudied phenomenon in civil wars. I theorize the role of threats to survival in generating cooperation among armed actors, arguing that reprieves from violent elimination by the enemy cause cooperation to break down. The assurance of survival pushes an armed actor to fight its allies, with whom it competes most intensely for political support, with an eye towards increasing its political power in the eventual post-war period. At the same time, the war continues to tie the fratricidal groups together as meaningful allies. This theory differs from existing explanations of the fragmentation of armed actors and shifting alliances within civil wars, which view threats to survival as generating break-downs in cooperation. Here, the opposite is the case: cooperation between allies breaks down when survival is secured. I test the theory using data on armed group interactions in the Syrian civil war. I also use structured comparisons within the subnational research design to check theoretical implications concerning the effects of variation in the cleavage separating competing allies (e.g. ideological versus identity-based; local versus regional or national) and in the source of the reprieve from violent elimination (e.g. infusion of resources from an external patron versus unexpected military gains). These comparisons allow for a nuanced understanding of conflict and cooperation between Islamist groups and the Free Syrian Army, umbrella groups and their component organizations, and Arab and Kurdish armed groups.