Splitting on same feature repetitively

I have a large dataset with 200 features and ~10m samples. I am having trouble because I keep getting decision trees which split on the same feature over and over (please see picture).

What can I do to prevent XGBoost from doing this? Or does this somehow make sense? Regardless of the data, the splits don’t make sense… because how can you have a feature that interacts with itself?

I don’t see why a feature should not interact with itself.

Think of the toy example where the single feature is a percentage score and the class output is the alphabet grade (A, B, C, D, F). Then you’d have a classifier that fits 0 <= X < 60, 60 <= X < 70, 70 <= X < 80, 80 <= X < 90 and X >= 90, requiring splits that splits on the same feature over and over again.

Admittedly this is a contrived example, but my point is that whenever it makes sense to fit a range [x1, x2] on a feature, you’d need to split on the same feature twice.