Due to the brain’s limited energy storage capacity, a robust blood supply via the vasculature is crucial to provide oxygen and nutrients. The cortical vasculature is characterized by its hierarchical structure consisting of the planar pial network at the cortical surface, followed by penetrating arterioles and venules. These vascular trees are connected by the highly interconnected capillary bed, which is the key location for oxygen and nutrient discharge. Evidence is increasing that during ageing and disease microvascular alterations might impair energy supply locally. Possible alterations are occlusions of individual capillaries or the dilation of subsets of capillaries in response to pericyte loss. In the first part, we employ our bi-phasic blood flow model to investigate the impact of the mentioned alterations on local perfusion characteristics in realistic microvascular networks (MVNs). In the second part, our focus is on large scale disturbances caused by the occlusion of the middle cerebral artery (MCA), i.e. one of the brain's main feeding vessels. Here, we employ an inverse modeling approach that allows us to incorporate in vivo data and thus to directly align our study to in vivo experiments. We use this framework to investigate the role of anastomoses between pial feeding regions, so-called collaterals, on perfusion changes in response to stroke. Our results show that both single capillary occlusion and the dilation of ~15 capillaries cause a local redistribution of blood flow. The precise response is governed by the local vascular topology. Interestingly, the most robust local topology is also the most common one. This indicates that the capillary bed offers an inherent robustness to single capillary occlusions. In case of multi-capillary dilations, we note locally increased perfusion heterogeneity. This might hinder flow homogenization as observed during neuronal activation. As the extent of collaterals is a key factor for the tissue volume damaged during stroke, we vary the number of collaterals in four semi-realistic MVNs. Both the number and the dilation of collaterals increase the redistribution of flow into the territory originally fed by the MCA. This redistribution comes at the cost of reduced perfusion of other areas. Importantly, this local decrease in perfusion is only captured if flow changes in penetrating vessels are monitored and might be missed if imaging techniques exclusively focus on vessels at the cortical surface.