* Split `reconnect_failed_client()` logic

* Reorganize block checking logic as state machine

This way we'll be able to save the state in case of a failure