CH 1.32 used blocking io operations. That resulted in code easy to write/maintain, but caused performance issues (reading data from source prevents writing data to take place in parallel).
Introduce async io in 1.40 to reduce the delays and speed up copying.
In 1.40 we expect to have a basic async io - i.e. operating on a single file at a time. In the future it should be possible to handle multiple files in that manner.