Building on the methodology for multi-scale load prediction and component failure analysis from part 1, this paper completes the quantitative reliability framework by presenting the systemic failure propagation analysis. Vessel burst effects are modelled using Bayesian networks. Overpressure waves are calculated using trinitrotoluene (TNT) equivalent methods, while heat fluxes from hydrogen fireballs are predicted using established correlations. The multi-energy model accounts for gas cloud deflagration characteristics during delayed ignition. Results demonstrate that almost all burst events lead to aircraft crash, as overpressure and heat flux exceed structural thresholds. Crash probability is calculated as a function of laminate thickness elevation factor, showing that burst-related causes dominate at low values while thermal pressure relief device (TPRD) activation creates an irreducible lower bound at high values. Sensitivity analysis reveals significant influence of material scatter on required vessel mass, while random fibre angle deviations show no significant influence for thin-ply vessels. Integration with part 1 enables design optimisation based on crash probability rather than burst probability alone.