As a number of folks have pointed out it's not the magnetic particles - rust is pretty stable; they read a 60 year old RAMAC at the CHM every week or so.
And it's not likely the binder, fully cured epoxies are also very stable. Note disks do not have the tension problem that tape has. And it's not likely the substrate, again Mylar is very stable My guess it is both contamination and wear of the head. After all, this is contact recording. Tom