Dataflow - distinct got stuck with 3MB of data and kills throughput

Question

I have python dataflow job that reads data from pubsub, extract some user information, window them into 5min timeframe, distinct and then do some more processing.

<