def withBackoff[T](minBackoff: FiniteDuration, maxBackoff: FiniteDuration, randomFactor: Double)(sinkFactory: () ⇒ Sink[T, _]): Sink[T, NotUsed] def withBackoff[T](minBackoff: FiniteDuration, maxBackoff: FiniteDuration, randomFactor: Double, maxRestarts: Int)(sinkFactory: () ⇒ Sink[T, _]): Sink[T, NotUsed]
This SinkSink will never cancel, since cancellation by the wrapped SinkSink is always handled by restarting it. The wrapped SinkSink can however be completed by feeding a completion or error into this SinkSink. When that happens, the SinkSink, if currently running, will terminate and will not be restarted. This can be triggered simply by the upstream completing, or externally by introducing a KillSwitch right before this SinkSink in the graph.
The restart process is inherently lossy, since there is no coordination between cancelling and the sending of messages. When the wrapped SinkSink does cancel, this SinkSink will backpressure, however any elements already sent may have been lost.