Via Ars Technica I found the linked document which explains with technical rigor what went wrong with the OPERA speed measurements – there was more than just one technical issue, really difficult to fully diagnose (and trace back). For me, incidentally, the mechanisms that explains the delay via incomplete light fibre connection, ties right in with VEP (visual evoked potential) delays when the retinal stimulus has reduced contrast, e.g. by wrong glasses / wrong adaptation. Though, of course, on a time scale of ms rather than ns.