> I think the most likely explanation is that this was a hardware error caused by a cosmic ray or the like, rather than a software bug. It's just very bad luck

> I think the most likely explanation is that this was a hardware error caused by a cosmic ray or the like, rather than a software bug. It's just very bad luck :-( These sorts of cosmic bitflips have been exploited as a security error for some time. See bitsquatting for domain names. ECC helps and so does software error detection on critical data in memory or on disk. The problem is always that input data should not be relied upon as correct. Do not trust data. It's a difficult mind shift, but it needs to be done, or programs will continue to mysteriously fail when data becomes corrupted.