On average, there are 64 mutations per generation in the human genome. Is this constant, or can we expect variation in the number of mutations?
Mutations in this context are being treated as discrete count data (there either happen or they don't). So the number of mutations per person in a given generation should form a discrete probability distribution. Because they are a single count variable bounded at zero (can't have negative counts) and finite (genomes don't go on forever), a Poisson distribution(pwɑːsɒn/) is probably the expected best fit. A Poisson distribution can sometimes be approximated by a normal distribution when the Poisson count mean (λ) is large (some say λ>20, some say λ>100).
Note: there are 60ish substitutions per generation, on average. There are probably many times that number including all mutations (e.g. microsatellites, centromeric satellite repeats, etc). There is almost by definition a telomeric mutation on every chromosome end at every cell division, which alone is comparable to substitution number per generation.
Figure 1d of this paper suggests that the distribution of substitutions alone is log-normal in cancer, as suggested by the commenter.
I am not aware of a reference for distribution of other mutations.