Why DATE(updated_at) Broke Timezones and Killed PostgreSQL Index Performance

I recently resolved two separate production bugs that turned out to share the same root cause — a seemingly innocent DATE() call wrapped around a timestamptz column. What made it worse: the pattern ha