Bentrold,
CDR discs use a dye that undergoes a chemical reaction when the write beam hits it. This reaction leaves a spot on the disc with lower reflectivity that the reading beam and detector can differentiate from the unwritten portions. The reaction is not reversible so once a pit is written it stays written.
You can check the article
here for a brief description of the CDRW technology. Essentially it involves melting a thin metal alloy layer. If you zap it with a strong enough pulse it will cool quickly and end up in an amorphous phase (disorderd) that has lower reflectivity. If you anneal the disc (heat all the metal layer up to just below the melting point) you can erase it by restoring the metal layer to crystalline form which has higher reflectivity.
The article claims a thousand rewrite cycles are possible with some manufacturers claiming even more.
I would guess the failure mechanism is some physical damage done to the disc by repeated thermal cycling that makes it impossible to detect the change in reflectivity through the damage. I doubt there is any way to easily repair this, certainly there is no econcomical way to do it (or you'd see people trying to reprocess these discs for money).