Boy you're doing a lot of listening lately ;~).
All sensors coming out of fabrication during manufacturing are tested. Each maker has an allowable tolerance for "dead pixels," basically photosites that are inoperative. Sensors with fewer than that number have those photosites identified and those are then mapped out in software later in the process of building the camera. Sensors with more defects than the tolerance limit are thrown away.
So what's this re-mapping thing? Basically all cameras have a function in their firmware that allows the manufacturer to specify that a photosite be removed from the demosaic and a substitute value (calculated from neighbors) be submitted in its place. Because Bayer systems are already doing such neighbor value calculations to produce pixels, usually you can't see the impact of this (indeed, I doubt you'd ever see it unless there were multiple adjacent defects). A few cameras—Olympus is notable for this—can perform tests and re-map pixels via functions in their menu system. Nikon cameras (except for the Nikon 1 bodies) require that the camera return to Nikon for re-mapping, should you discover you need it (photosites can die in use, though it's rare).
A related issue is hot pixels. Hot pixels are photosites that are producing false bright values. The reasons they might do so are numerous, but heat is often part of the cause. Hot pixels, like dead pixels, can be remapped. Generally I don't recommend that unless you have a persistent hot pixel in normal shooting (low ISO, high shutter speeds). Nikon, however, won't remap hot pixels unless you have more than a minimum number and these occur at shutter speeds faster than (I think) 1/4 second at low ISO values. If you shoot raw, you can remap hot pixels automatically by just using the right raw converter. Adobe's converters (Lightroom, ACR) detect and map out hot pixels automatically, for instance.