Replication Upgrade process is broken

When a replication pipeline is upgraded, the previous state which is associated with an Pipeline identifier is lost. When if a replication pipeline is upgraded it loses its state/checkpoint and starts all over again.
This is because during upgrade we do not have the previous pipeline identifier (generation).

But Upgrade from any lower version(lower than 6.6.0 ) to 6.7.x+ is handled and upgrade works FINE. This is because we are migrating checkpoint info from GCS to CloudSql.

But for patch upgrade, for example, 6.7.0 → 6.7.1 will also lose state/checkpoint and start over.

This was a known issue previously as well ( lower than 6.6), but since it was in GCS , we used to guide the customer to edit the state/checkpoint file manually and restore the State.