Cleanup `Thread.sleep` in unit tests

[sijie]

Problem

Currently we are using a lot of Thread.sleep in the unit tests, waiting for some tasks to complete execution. This pattern is very problematic, which it makes the tests unstable and non-deterministic.

Solution

We should try to clean up the usages of Thread.sleep in the unit tests. This is a master ticket for tracking this effort.

[eolivelli]

This is related to #751