I've been playing around with slice_sample on grouped data frames for bootstrapping. The documentation states that if you provide a prop>1 or n> # of rows within a group, it will only sample the number of rows within a group. If you have replace=TRUE, that's not respected and you get as many samples per group as you request. I thought it was going to be a slick way to handle resampling from groups with different numbers of observations