A half-life of 10,000 years means the fraction (0 to 1) of healthy bonds at year t can be modeled by (1/2)^(t/10000). To have 1% bonds broken, it would take an expected 145 years.
EDIT: Fixed the numbers, my bad!
>>> log(0.99, (1/2)**(1/10000))
144.99569695122509[1] It may be a bit worse in the long run, since if many bonds break, the higher order DNA structure gets damaged and that may change the rate of decay experienced by individual bonds (?).
If indeed 1% break after 15 years, then 10% (ignoring compounding) break after 150 years, and 50% after 750 years. That doesn’t square with a half life of 10,000.