I just came across a similar breakdown of Apache Pinot vs Clickhouse vs Druid at https://startree.ai/resources/a-tale-of-three-real-time-olap...
We're experimenting with Pinot for it's ability to do low latency query on streaming data at high concurrency, tight integration with Kafka, and the ease of which it handles upserts.