We investigate how mineral dissolution reshapes flow pathways and solute transport in three-dimensional discrete fracture networks (DFN) using a computationally efficient graph-based reactive transport model. The DFNs are inspired by field-site observations of fractured carbonate and represent realistic connectivity and structural heterogeneity. Flow is simulated with the Reynolds equation, and dissolution follows first-order kinetics with diffusive limitations captured through an effective mass-transfer coefficient. By systematically varying two key dimensionless parameters, the effective Damköhler number (Da), governing reaction versus advection rates, and a diffusion-controlled reaction parameter (G), analogous to the Thiele modulus, distinct flow channelization regimes emerge: mildly channelized at low Da, highly channelized at intermediate Da, and extreme wormhole formation at high Da and low G. Eulerian and Lagrangian analyses, including breakthrough curves, particle tortuosity, dispersivity, and flow channeling indicators quantitatively characterize the progression of dissolution-driven network restructuring. Across all regimes, the initial fracture heterogeneity imposes persistent heterogeneity. The results underscore how the balance of initial heterogeneity, advection, reaction, and diffusion critically shapes subsurface flow pathways, with implications for applications ranging from groundwater remediation to enhanced geothermal systems.