Launch operations and reliability readiness

Problem

The inspect-profiler runs database profiling, semantic typing, and generates inspector reports for analysts, but has no operational readiness for production use. There is no telemetry on profiling job durations, failure rates by warehouse type, or semantic type detection accuracy. If a profiling run times out against a large Snowflake warehouse or produces incorrect type classifications, nothing alerts the team. Support ownership for analyst-facing inspector issues — stale profiles, wrong column types, failed report generation — is unassigned. Without an incident playbook, profiling failures at launch would leave analysts with missing or misleading data quality insights and no escalation path.

Context

Possible Solutions

Plan

Implementation Progress

Review Feedback

• Review cleared