The iOS crash that hid for weeks—here's how they finally found it.

A large-scale consumer iOS app was plagued by an elusive, intermittent SIGBUS crash that affected only some users and couldn't be reproduced locally. The crash reports pointed to Foundation's NSAttributedString handling HTML imports, but the stack traces were vague, leaving developers to hunt through a wide search space of object lifetime, concurrency issues, and SwiftUI view code without a clear path to follow. The problem resisted all standard debugging approaches because the failure occurred probabilistically under specific timing conditions.

This debugging journey demonstrates how subtle threading violations can hide in plain sight: HTML import through NSAttributedString secretly requires the main thread, a requirement easily missed when the call sits inside asynchronous screen preparation code. The solution involved creating controlled experiments to isolate the fault, ultimately fixing it with a small code change that enforced the threading contract through a shared main-actor abstraction. The key takeaway is that one reproducible device can turn speculative debugging into engineering work, and that a lower crash rate doesn't prove a fix—it might just change the timing of a lurking concurrency defect.