Actual device failure rates are quite low. A large study published in 2005 by the University of Michigan found that 5.1 percent of their surgeries are revision surgeries, that is, to replace an existing device and less than half of those were for internal electrode failure, which means over several years at a very large implant center, only 2.5 percent of their annual procedures are to replace devices where the electrode failed. There is an abstract summarizing this paper available on pubmed.org
Keep in mind that those numbers are for devices that were manufactured over many years, and the technology keeps improving. The failure rates cited at the recent CI 2007 conference in Charlotte for the newest generation of CIs is under 1 percent when you eliminate trauma as the reason for the device failure.
That's not to say it never happens. Of the 600 appeals we handled last year, I think 15 were to replace devices with internal failures. But the original poster is correct that it is REALLY important to distinguish reasons for bad outcomes, as there are many reasons for bad outcomes other than the device itself.
Someone mentioned etiology, specifically ossification, is occasionally associated with less than optimal outcomes. This is true, a couple of other etiologies associated with below average outcomes are Waardenburg Syndrome and Auditory Neuropathy. Waardenburg has to do with the density of neural ganglion which is really unpredictable in those patients, and Auditory Neuropathy results have to do with white noise being generated in the unimplanted ear -- assuming the patient doesn't have a bilateral implant, which obviously solves this problem. Note that there are people with Waardenburg and AN who get perfectly good results with either one or two implants, just as there are those with meningitis who get great results it is just that these are a couple of sources of HI that I've heard that can be associated with less than perfect CI results.
Sheri