481: Dev Horror Stories

What It Covers

Sally and Joelle share developer horror stories from their careers, covering production database disasters, data corruption bugs, time zone issues, debugging challenges, and the scariest things encountered when joining new projects.

Key Questions Answered

• •Database Recovery Strategy: When production data is lost, combine recent backups with transaction logs to rebuild state. After dropping a production database, recreating every transaction from logs since the last backup restored all data despite the backup being 24 hours old.
• •Case-Insensitive Search Performance: Using Postgres ILIKE for case-insensitive matching bypasses existing indexes, causing severe slowdowns at scale. Switching to citext column type maintains case-insensitive searching while preserving index performance, solving import bottlenecks when processing 10,000 records simultaneously.
• •Multi-Column Index Ordering: Index column order must match query WHERE clause column order exactly in Postgres. An index on name and location only works efficiently if queries use WHERE name AND location, not WHERE location AND name, making query planning critical.
• •Time Precision Bugs: Timestamp precision mismatches between Postgres and Ruby create flaky tests when subsecond values are converted to strings. Bugs appeared only when decimal portions were divisible by eight due to octal interpretation, requiring careful handling of time data transformations.