What does it mean that oobPermute​dPredictor​Importance is "Randomly permuting the observations of predictor variable"?

I am using oobPermutedPredictorImportance to determine feature importance with random forest regression on 10-15 features on a relatively small number of observations (~100).

When reading MathWorks Help, section "More about" it says the following about the method of permuting the features:

...

For each predictor variable xj, j ∊ st:

1. Randomly permute the observations of xj.

...

I want to know if this random permutation is limited in the range of my input data.

Example: I have a feature v "fluid velocity" in range from 1...15 m/s which is used when building a RF model. Phisically it makes no sense to go outside of this range. Do the random permutations of this predictor variable stay inside this range?