Practical recommendations
If you’re building Databricks Apps:

Optimize for iteration first

• Start with external app + Databricks backend if UX-heavy.
• Or keep apps thin and logic in tables/models

Decouple UI from compute

• Use SQL Warehouses or Model Serving
• Avoid long-lived Spark sessions for apps

Treat apps like products, not notebooks

• CI/CD from day one
• Version everything
• Separate dev/test/prod workspaces if possible

Invest early in local dev parity

• Docker images matching Databricks runtimes
• Mock UC permissions
• Config-driven environments

Plan for scale & multi-user behavior

• Stateless app design where possible
• Explicit session handling
• Avoid global variables