I agree completely with the Microchip reply...
However, doesn't it looks too much like a 'canned statement'?
I wouldn't doubt that, all other things being equal, a PIC running at 20Mhz would fail sooner than that exact same PIC running at 4Mhz.
But how much sooner? What kind of time scale are we talking about here? Would any of us, or our offspring be around to mark the time at which one or the other failed? Kind of like measuring how long an EEPROM will hold data. EEPROMs haven't been around for the 100's of years it would take for one to supposedly degrade and lose data.