It has definitely been debunked. Don't triple down on this.
Do you have a background in statistics? Machine Learning? I'm not an expert in either (Computer Engineering Major with a minor in BioChem) but I do work with people that apply Benford on
continous datasets (election votes are discrete sets by the way
) that have some form of randomness involved in said dataset. We use it as
one indicator to determine if data from our devices are producing valid data to ensure that a device isn't broken. It's not a definitive statistical model and it needs to be paired with other ones.
anyways I'll let the mathematicians speak
http://www-personal.umich.edu/~wmebane/inapB.pdf (2011)
http://www-personal.umich.edu/~wmebane/inapB.pdf (2020)