Relation between table manger and s3 (object store)

I have a question about loki’s table manager.

I have installed loki via loki-distributed helm chart. and the table manager is running as a pod (replicaset)

According to the documentation Table manager | Grafana Loki documentation , the table manager creates a table to store chunk data. Where exactly is it stored?

Looking at the architecture, the table manager is connected to the index store (= dynamodb in our case).
Table manager is not a statefulset using pv and the capacity does not increase much, so I do not know where the chunk data is stored. or does it just mean metadata?

Also, we are using amazon s3 to store actual logs (chunk data).
According to the documentation, object storage such as s3 is not supported by the table manager, so s3 lifecycle policy must be applied for log retention.
So, when I set the policy to s3 as 365d and set the retention_period / max_look_back_period to 24h, I could not retrieve data older than 24 hours in Grafana Loki dashboard.

Does the table manager only fetch data from s3 according to the retention setting? I want to get confirmation whether the table manager is holding the data directly.

Also, if the retention period of the table manager is set as long as 360d, I wonder if there may be capacity or performance issues.

If you are running recent versions of Loki, you should use compactor instead of table manager. See Retention | Grafana Loki documentation.