Why does the results of the bootstrapping methods differs if it is being used the same seed?

Question

I want to generate 95% confidence intervals from the R2 of a linear model. While developing the code and using the same seed for both approaches, I figured it out that doing the