MTBF and warranties are only marginally and periphally about engineering - their real nature is marketing.
The few broadly-based scholarly reviews suggest annual replacement rates (AFRs) are in the 2-4% range for BOTH consumer and enterprise drives (with a 'fat tail' of lemons going out above 15%!). (I will blur here the distinction between replacement rates and failure rates). Moreover, the failure pattern is far better represented as a Weibull or gamma distribution than a Poisson one! In short, real in-the-field MTBFs are in the 200,000 to 400,000 range for all drives. MTBF numbers above 500,000 are, to put it charitably, a grossly optimistic artifact of flawed testing methods or - and here's the rub! - of how the testing information is (mis-)represented!
Warranties are largely (not entirely, just 90% :-) marketing hype. I could take EXACTLY the same drive and offer it with a 2-year or 3-year warranty (i.e., nothing to do with the real underlying drive construction or failure rate). It would only incrementally cost - in the typical case! - 3% of the drive cost to offer the longer warranty. (And that's assuming - falsely! - that all failed drives will be returned as a warranty claim and, even of those that are returned, the manufacturer does not disallow any of the returns as invalid!)
We're talking a cost of perhaps $3 a drive on a 2-TB drive (more like $1.50 if not all failed drives are returned and not all returned drives are considered valid warranty returns).
$1.50 is peanuts if I can increase sales and market share with my 3-year warranty or even - and this is the usual case - charge, say, a $5.00 premium for my 'better' HD with its 3-year warranty over the 'inferior' one with a 2-year warranty.
Anyone who has ever bought an 'extended warranty' from a retailer soon learns that these are overpriced cons. Well, manufaturers' warranties are no better. Mostly marketing hype (where mostly > 90% :-).
Regards,