This may be true for one specific site, but when you look at many sites, you expect to see more frequent events. This is a mathematical fact and is explained by probabilities and the binomial distribution. Hence, this statement is slightly misleading, since similar events could have happened in other locations, and when applying analysis of return intervals, you need to take into consideration the number of possible multiple sites [if n=number of sites, then a binomial distribution Pr(X>4) would be more appropriate, where p=1/500, sample size=n, and X is the number of events]. However, this makes also the situation more dramatic, as a slight shift in the probability of an extreme event will pan out to cause more events somewhere in these multiple sites.