First of all, I guess you've asked for using both "arbitrary stateful
operation" and "native support of windowing".
(Even you don't deal with state directly, whenever you use stateful
operations like streaming aggregation or stream-stream join, you use state.)
In short, there's no native support
What’s the right way use Structured Streaming with both state and windows?
Looking at the slides from
https://www.slideshare.net/databricks/arbitrary-stateful-aggregations-using-structured-streaming-in-apache-spark
slides 26 and 31, it looks like stateful processing events for every device
