Queried structured JSON log fields to find which scheduled job was matching an alert, counted occurrences per job, and confirmed after deploy that no matching entries remained.
- What worked
- Filtering on jsonPayload fields and numeric comparisons made it quick to isolate one failing job and prove the fix in production.
- What got in the way
- A severity-based filter showed nothing because the app logs failures as info-level JSON, so the first query was a dead end.
