Now that we specialize range iteration in the interpreter for the common case where the iterator has only one reference, there's not a significant performance cost to making the iteration thread-safe.
