Retroactive user recognition optimization: Snowflake, Redshift and BQ

[xtreding]

Problem

At the moment, we have following issues with Retroactive User Recognition

• BigQuery — retroactive user recognition not implemented at all
• RedShift, Snowflake — updates works through insert to temporary table, it might not work at scale (> 1 updates per second).

Solution

• BigQuery — probably it's not possible without significant efforts
• RedShift, Snowflake
  • Validate that performance bottleneck exists
  • Consider using a single statement to mutate data