The Silent NullsPipeline · medium · ~12 minA mobile SDK release on Tuesday stopped setting session_id for ~30% of events. Ingestion accepted them (the column is nullable), every downstream join quietly dropped the null-keyed rows, and by Friday three teams had shipped decisions based on dashboards showing a 30% 'engagement drop'. The pipeline was green for 96 straight hours.Build the quality system that makes this impossible to miss in under an hour.Evidence

Metrics

Evidence

Config

Checkpoints

What's the correct reading of the evidence?

The quality architecture that catches this in under an hour?

Where do checks belong (ingestion vs transformation vs BI), and what's the playbook when the gate fires? Interview-style.