Use case

• Streaming queries (that we don't have).
• Filling aggregated materialized views.

Describe the solution you'd like

Some sections in query to specify:

• do flushing for every N groups or every N of input rows or every amount of real time passed (e.g. every second) or on some other criteria based on data;
• output result on current part of data and start processing next part of data from scratch or output intermediate accumulated result.