It's the fact that it's silent that is the problem. Writes fail under certain circumstances, and that's acceptable, but when a failed write perhaps affects other data silently and now your system reports healthy but you don't know if it really is... That's just .. ouch.