I am interested to hear how others are using the Dagster DuckDB integration. It seems really neat to be able to use DuckDB as an IO Manager -- but how are people persisting the DuckDB across runs when using Kubernetes or ECS? I am using Dagster Cloud in the Hybrid mode with ECS and whenever I start a new run there is a new ECS task created to host the run. I am assuming the DuckDB process would run as part of this task. It seems awesome to be able to use DuckDB as an IO Manager between OPs -- but what happens if the run finishes and I want to later re-run some of the Ops? The DuckDB process and data would be cleaned up with the ECS Task? Is anyone using DuckDB in a way that makes it more persistent?