We encountered a segmentation fault in mongod during startup recovery when a WiredTiger journal file (journal/WiredTigerLog.*) was partially corrupted. The segmentation fault occurs in __wt_log_scan during the wiredtiger_open recovery process.

More critically, each restart attempt in this corrupted state creates a new ~100MB journal file, even though the process eventually crashes due to the corruption. If the service is automatically restarted (e.g., by a process manager) many times (e.g., 10,000 attempts), it can generate approximately 1TB of journal files, potentially filling the data disk and causing operational disruption.