The study of percolation phenomena has various applications in natural sciences and, therefore, efficient algorithms have been developed to estimate the corresponding percolation thresholds. For instance, this applies to the widely-used bond-site percolation model for which the Newman-Ziff algorithm enables an efficient simulation. Here, we consider several non-standard percolation models that have applications in measurement-based photonic quantum computing with graph states. We focus on prominent architectures where large-scale graph states are created by fusion networks connecting many small resource states. We investigate percolation models that provide an estimate of the tolerance to photon loss in such systems and we develop efficient algorithms to analyze them through modifications of the Newman-Ziff algorithm. We consider non-adaptive fusion networks with all fusions being performed at once, and adaptive ones where fusions are repeated conditioned on the outcome of previous fusion attempts. We demonstrate our algorithms by using them to characterize several fusion networks and provide the corresponding source code.