The new test has a flakiness factor on slow platforms, where the close
on the sending connection races with the last index message, potentially
messing up the count. This adds a wait to ensure that all sent messages
are received, or the test will eventually fail with a timeout.