Yes, xarg can run in parallel. The problem is that the find operations are always bottlenecked by I/O of the zpool. So whether you have a single task running in series or an arbitrary number of tasks running in parallel won't matter, except you'll potentially lose the efficiency of running in series. I'd wager that for most users, if they don't have 8GB of RAM, will have a metadata ARC size that is sufficient to prevent any significant difference between serial and parallel operations. But, if you don't have enough RAM, the parallel operation will be noticable (maybe even a factor or two longer) than in series.
Short and dirty.. serial is probably the better way to go for everyone across the board.
Short and dirty.. serial is probably the better way to go for everyone across the board.