I made some tests a while ago about how often herb generations occur (level 50 wizard in UD:1).
Total of 7889 generations were made (pasting a lot of wait + eat commands in sage and parsing debug log afterwards).
I'm not sure if it's exponentially distributed or not. It looks quite a lot like exponential function and the best correlating function that I could find was exponential function, except that generations in < 5 turns seems to have way too small probability for it to be an exponential (if it were exponential function, 1 turn generations should be most common, 2 turns second most common and so on).
The lowest number of turns between generations were 1 (occured on consecutive turns) and it happened only 35 times. The highest 5 were 1301, 1276, 1049, 1011 and 999 turns so there doesn't seem to be upper limit.
The average number of turns between 2 generations were 134 turns and median was 92 turns (half of the generations occured in less or equal than 92 turns and half in more or equal than 92 turns).
If someone is interested in looking the data collected, it can be found here:
http://www.students.tut.fi/~maki36/a...enerations.txt
There are only numbers in the file and every number means how many turns it took for next generation to occur.