Generated points are not reduced when sampling spacing increases

Point sampling filter is for generating additional points. I don’t think it can reduce the number of points - see documentation.

Poisson dusk sampler can also add random samples (see Selecting random points of a PolyData (Stratified sampling)) but I don’t think it can reduce the number of points (simplify the mesh) either.

There are decimation filters in VTK that you can use to reduce the number of points, but the result is not uniform, but it will strive to preserve all details with less points, so sampling will be sparse in flat regions.

Unfortunately, VTK currently does not have a filter for uniform resampling, but you can use the ACVD algorithm for that, for example using the pyacvd package.