The $400k Bug That Only Happened in Production

Lucas and Luna dive into a notorious production-only bug from a mid-size fintech company: a memory corruption issue that triggered random data loss in user transactions. The bug only manifested under real load and took three weeks to root-cause. They break down the debugging process, the tools used (GDB, address sanitizer, core dumps), and the engineering lesson about Heisenbugs and timing-dependent memory corruption. Specific details include the bug's signature — a race condition in a custom memory allocator that zeroed out memory only when allocation patterns hit a specific threshold. The episode also touches on how the company built a regression test by replaying production traffic at lower concurrency to reproduce the issue. #ProductionBug #Heisenbug #MemoryCorruption #RaceCondition #GDB #AddressSanitizer #CoreDump #Debugging #Fintech #SoftwareEngineering #MemoryAllocator #RegressionTesting #ProductionTrafficReplay #ConcurrencyBug #Technology #FexingoBusiness #SoftwareEngineeringPodcast #CodeDebugging Keep every episode free: buymeacoffee.com/fexingo