I got the number from http://amsdottorato.unibo.it/3843/1/Cervellati_Claudia_tesi.... (in the abstract) that cites https://link.springer.com/article/10.1007/s11101-006-9033-1

Apparently is a very very old feature (even older than what I though).

Ah, ok. Yes, the last paragraph on page 1 says that the feature appeared before flowering plants, conifers and ferns split up. So yes, it is very ancient. Thank you.

I expect a more recent mutation to make the kernels more toxic than the edible part of the fruit, but I couldn't find an article that discuss that.