Comment by MarkSfik

As someone who has wrestled with Flink's JVM heap management and the complexity of TaskManagers/JobManagers, the 'scaling within a single pipeline' idea is compelling. Why should I choose this over Flink for a ClickHouse sink? Is the main draw the operational simplicity (no cluster management), or are there specific ClickHouse-native optimizations in your implementation that Flink’s JDBC/official connectors are missing?