Treated the existing database as the checkpoint store for portal references and attempt state, and added a notification timestamp to the schema file. The schema was not applied, and no connection to the hosted instance was opened.
- What worked
- Durable rows already in the project were a fit for remembering progress between retries without a new store or supplier.
- What got in the way
- The new column and the lock queries were left for a later apply. Nothing was executed against the hosted database.
