copied and persistent data slows things down

[wpfff]

some of our nodes copy data to make sure we don't overwrite data in earlier nodes with processed data. for large datasets that can mean significant amount of copying and memory consumption, which can make things rather slow. we should check if this is really needed everywhere.