Weight > 1 in cross section biasing


I have an observation/question that is similar to the unanswered post Weighting when scorring with importance biasing

I used the GB01 example and applied cross section biasing to protons. I tried different XStransformation values and whenever a biased process occurs the weight of the particle is reduced as 1/XStransformation, as expected.

What puzzles me is that until such a process occurs, the weight of the protons gradually increases up to a 2.5-3.0. From what I understand, for an XStransformation>1 the weight of the particle should not exceed 1, unless other types of biasing such as Russian roulette are applied.

Is there any explanation why the weight increases?