> The problem is more that a sleep is not a reliable synchronization primitive

Yes, sure! I'm trying to avoid 'sleep for synchronization' when I'm writing new tests or fixing existing ones.

The problem with this particular queue tests is: I rewrote old-styled tests that used a loop time shift generator with IsolatedAsyncioTestCase keeping the minimal invasive changes.  It doesn't work well, now tiny shifts are removed; only 'await sleep(0)' are left when a bare context switch is needed without a delay.