Real reliability defects, the incidents they caused, and what a Tomosu scan found. Each study runs Tomosu blind on the code as it shipped, with no bug report, patch or hint, and compares the result with the fix engineers eventually made.
3,000 users exposed it. Could the code have told us first?
A load test left PostgreSQL connections stuck idle in transaction until the pool ran dry. A blind scan of the pre-fix token_scoping.py named the same unsafe session cleanup, judged against a criterion locked before scanning.
A memory leak users found first
Reported in 2018, fixed in 2026. A blind scan of Celery's pre-fix events/state.py found the exact unbounded-growth failure mode and recommended the same bound the maintainers shipped.
Want to see what Tomosu finds in your code? Run a free PRI scan on any GitHub repository.