Comment by arjit_jain

Great catch! Importance splitting is indeed a useful simulation technique for low probability events. The application of splitting based methods on bayesian models for extreme sample generation can be an interesting future work.