[El Gato]

Actually, over 30 years, something with a 0.5% failure rate increases to about 14%

True, if the failure rate is 0.5% per year, although they didn't say that. However 0.5% per year is one per 200 years, so that is likely what they meant. Note that using your methodology, you get 63% probability in 200 years.

[supercat]

True, if the failure rate is 0.5% per year, although they didn't say that. However 0.5% per year is one per 200 years, so that is likely what they meant. Note that using your methodology, you get 63% probability in 200 years.

Which is still only an average of once every 200 years.