Distributed streams can get stuck in infinite loop, wasting CPU if a new toplogy is never installed. This affects all cache modes (scattered, dist and repl).

We should add a wait for the next topology when we complete a loop and not all segments are completed.