Somewhat tangentially, I'd like to point out that simply bringing up relative frequencies of different types of events in discussion doesn't make one a crypto-frequentist -- the Bayesian approach doesn't bar relative frequencies from consideration. In contrast, frequentism does deprecate the use of mathematical probability as a model or representation of degrees of belief/plausibility.

Er, no, they're called Dynamic Bayes Nets. And there are no known unique events relative to the fundamental laws of physics; those would be termed "miracles". Physics repeats perfectly - there's no question of frequentism because there's no probabilities - and the higher-level complex events are one-time if you try to measure them precisely; Socrates died only once, etc.