When the offset SCN is quite old and the connector is restarted, the buffer is re-built from that offset SCN forward, and the issue is that this involves reading potentially dozens of logs.

Users often view this as "connector is stuck" and this creates a vicious cycle where they stop and restart the connector again and it continues to fall further and further behind.

We should see if we can process logs in smaller batches when using these mining strategies as a way to at least appear the connector is moving forward.