So for example we use trino to query delta tables, but trino is not great for ms latency.
So the alternative would be for example to do a dual sink and stream out to both a kafka topic and a delta table, and then handle low latency ms querying on topics rather than a delta table, but it would be nice to consolidate it all to delta tables if possible.
I did this in the past with apache druid to fulfil “realtime” dashboards but still keep the data available in delta tables for other BI and applications etc for long time storage since a topic is not meant to be a long time storage.