Yes, the bug in the code is actually available here:
Basically, the wrong parameter was being passed as the date to the function that calculates the sun rise/set times. This results in quite a bit of error and randomness. That randomness somehow worked in my local example and slipped through. This should be released today.