WD actually is using TDK heads right now with marvell channels. Also, why did they show a picture of a longitudinal head when everyone is using perpendicular now?
Oh, and about the reading data, actually with the advent of PRML none of your data is really "known" the bits are written so close together that it's hard to tell if you wrote 1010 or 1101, so they over sample the read then use a vitirbi probability tree to predict what the data actually was. The only reason why you dont see the errors come out of the drive when they occur is there are EXTENSIVE error correcting methods and re-allocations of bad sectors that occur. In fact in certain circumstances our drives (I wont say which company) can spend upwards of several seconds on one sector trying our darnedest to get the data back, if we cant, then it's time to re-allocate the sector.
We also do things to prevent errors, like back ground scans while the drive is idle, RAW (read after write), the fly height of the head is actually heater controlled so that over temperature and altitude the fly height stays the same (air density changes your fly height, the head is an air foil after all), skip write detect, there is also a shock sensor that will stop all writing when your drive shakes (in fact just rolling a screw driver handle over the top of the drive is enough to trigger these), a zero G sensor that detects free fall and retracts the heads so they dont get damaged (great for you Ipod).
Sorry if I made your feeling of "safe" data go away, but as you can see, ignorance is bliss because all that hard work usually pays off, it's not very that often you see a UDE (uncorrectable disk error)