Akka Stream GroupBy: we want to achieve ordered processing per substream + paralell processing between substreams

test.scala

/// Working version
val fixedThreadPool = ExecutionContext.fromExecutorService(
  Executors.newFixedThreadPool(10)
)

case class KafkaRecord(partition: Int, value: Int)

val records = (1 until 1000).map(i => KafkaRecord(i % 10, i))

val flow = Flow[KafkaRecord].mapAsync(1) { i =>
  Future {
    println(s"Processing $i")
    Thread.sleep(1000)
  }(fixedThreadPool)
}
.withAttributes(Attributes.inputBuffer(initial = 16, max = 16))
.async

val futDone = Source(records)
  .groupBy(10, _.partition)
  .via(flow)
  .mergeSubstreams
  .runWith(Sink.ignore)

Await.result(futDone, 2 minutes)