I'm trying to implement some logic in the pyflink table API using three separate kinesis data streams with shared keys. I'm not sure if I understand the concepts correctly or if what I'm trying to do is even possible... I have three kinesis data streams that I would like to apply some business logic to in real-time: • One of the tables should just have an unbounded accumulator running where I do a simple sum grouped by key. • The second table contains some events that should initialize and reset the accumulator, ie. starting the accumulation of the first table on some event and reset it to zero on other events. • The third table does lookups on the accumulator, getting enriched by the accumulation produced by the two other tables. I'm wondering if this kind of logic is possible using custom windowing? Or should I should use inserts into the accumulated table? Can aggregations be reset by events from another table? Maybe I'm framing this wrong entirely Thanks a lot :)