It creates more problems than it solves.

We replace this approach by a straightforward
synchronous one.

Now in the downlink, at frame f and subframe s
we use HARQ process (f * 10 + s) % 8.
This is similar to uplink scheduling.