Most likely, we're not passing this one because our implementation of tee is intentionally not conformant with the Streams spec. Our implementation is different because the standard approach uses too much memory. This is an area where the standard ought to be improved.

This will be a long-term task.