Could you run the benchmark by bootstrapping (average of repeated subsampling), instead of a straight-across performance score, and regain some leakage resistan

Could you run the benchmark by bootstrapping (average of repeated subsampling), instead of a straight-across performance score, and regain some leakage resistance that way? As well as a better simulation of "out of sample" data, at least for a little while.