As I read it, the bug was a minor issue. The bigger issue is the general false positive rate, which isn't a bug exactly, but more like a statistical modeling question - with millions of vocals, what's the appropriate way to characterize an activation as spurious?