Scala processor module executor trims messages

How to reproduce:

1. Run xd-singlenode (for which setting the Spark master URL to 'local' is a requirement). Use more than 1 worker thread. e.g. local[4]

2. Deploy the word-count example

3. Create a stream
stream create spark-streaming-word-count --definition "http | word-count | log" --deploy

4. Send data
xd:>http post --data "a b c d e f g"

xd:>http post --data "a b c"

5.Observe the result

2015-02-24 15:12:46,018 1.2.0.SNAP INFO Executor task launch worker-3 sink.spark-streaming-word-count - (e,1)
2015-02-24 15:12:46,018 1.2.0.SNAP INFO Executor task launch worker-1 sink.spark-streaming-word-count - (d,1)
2015-02-24 15:12:46,019 1.2.0.SNAP INFO Executor task launch worker-2 sink.spark-streaming-word-count - (b,1)
2015-02-24 15:12:46,020 1.2.0.SNAP INFO Executor task launch worker-1 sink.spark-streaming-word-count - (g,1)
2015-02-24 15:13:40,020 1.2.0.SNAP INFO Executor task launch worker-1 sink.spark-streaming-word-count - (a,1)
2015-02-24 15:13:40,020 1.2.0.SNAP INFO Executor task launch worker-2 sink.spark-streaming-word-count - (b,1)
2015-02-24 15:13:40,021 1.2.0.SNAP INFO Executor task launch worker-3 sink.spark-streaming-word-count - (c,1)

(the last three results are coming from the second invocation))

Note: there seems to be a correlation between the number of values emitted and the number of workers, as, in all the attempts, there aren't more values emitted than the number of workers.