Turn multiple Cassandra partitions into Akka stream

Question

I'm trying to develop a Source that turns a set of cassandra partitions into an Akka stream like this (I execute multiple cassandra queries through a bound statement, one for each partition and then I mapConcat the result into an iterable of DTOs):

Source<Result<JournalEntry>, NotUsed> src = Source.unfoldAsync(queryBag, s -> {

    BoundStatement bound = getBoundStatement(queryBag);

    // get out data from cassandra:
    Future<Result<JournalEntry>> page = 
        Future.apply(() -> journalMapper.map(session.execute(bound)), ec);

    Future<Optional<Pair<QueryBag,Result<JournalEntry>>>> next = page.map(r -> {
        Optional<Pair<QueryBag,Result<JournalEntry>>> opt;
        if (r.isExhausted()) {
            opt = Optional.empty();
        } else {
            opt = Optional.of(Pair.apply(queryBag.incrementPartitionId(), r));
        }
        return opt;
    }, ec);
    return FutureConverters.toJava(next);
});

Source<JournalEntry, NotUsed> concat = src.mapConcat(res -> () -> res.iterator());

// materialize concat and consume the stream

Is this a safe practice? In particular, I'm concerned about generating the iterable through mapConcat((res) -> () -> res.iterator()), as multiple calls of iterator() method of the resulting Iterable would get the same instance of iterator.

Besides, is this a sensible approach to iterate across different cassandra partitions or would it be more appropriate to use a GraphStage?