Currently, direct write-back from Lakebase PostgreSQL to Unity Catalog Delta tables is not the recommended pattern.

A few points that may help:

• “Sync Tables” are mainly designed for replication/read access scenarios. Databricks currently recommends avoiding direct writes to synced tables because consistency/governance behavior may not be guaranteed.

• If the goal is to move Delta data into PostgreSQL inside Lakebase, the better approach is usually:
- ETL/ELT pipelines using Databricks Jobs
- Structured Streaming
- COPY INTO / JDBC writes
- CDC-based pipelines

In practice, Databricks → PostgreSQL via JDBC or streaming is the safer supported architecture.

• For bidirectional workloads, many teams treat:
- Delta + Unity Catalog = analytical/storage layer
- PostgreSQL/Lakebase = transactional serving layer

Regarding your second question:

• Yes, PostgreSQL can be exposed in Unity Catalog using Lakehouse Federation (Foreign Catalog concept). PostgreSQL is one of the supported federated sources, so you can query PostgreSQL tables from UC without fully ingesting the data.

But federation mainly enables governed query access — it does not make PostgreSQL tables behave like native Delta tables with all UC-managed write semantics.

So today:
- Read/query interoperability → supported
- Native governed write-back from Lakebase PostgreSQL into UC Delta tables → limited/not the primary recommended pattern yet.