When would you use Streaming API versus a bulk data sync approach?

Streaming API (CometD / Platform Events / CDC): Real-time push of individual record changes or custom events to subscribed clients. Events are pushed as they occur, typically within seconds. Best for: real-time dashboards, instant notifications to external systems, triggering downstream workflows as soon as data changes. Limitations: 72-hour event retention, subscriber must manage Replay IDs, not suitable for initial full data loads or historical backfills.

Bulk Data Sync (Bulk API, scheduled SOQL export, ETL): Periodic batch extraction of large datasets. Use for: initial data loads into data warehouses, nightly sync to downstream systems, historical data backfills, data archiving. Tolerates latency (minutes to hours) in exchange for high throughput.

Hybrid approach: Use Bulk API for the initial full load (historical data), then switch to CDC/Streaming API for ongoing delta changes. This is the recommended pattern for data warehouse integrations (e.g., Salesforce → Snowflake via MuleSoft or a Salesforce change data capture listener).