https://www.history.com/news/mine-fire-burning-more-50-years...
Centralia is more of a weird natural disaster. It's essentially a slow-motion, underground forest fire. And, like most forest fires, while usually a human with a match is the proximal cause, if the human wasn't there, nature would still have supplied a spark eventually.
• What would the statistical yearly probability of a coal-seam fire be for seams we find, check up on, but never mine? (We know of a few seams in protected areas, but I don't think we know of enough to have any statistical power. Also, the natural catching-fire rate might be, say, one per thousand years, which would be a bit like the pitch-drop experiment in its requirements, yet would still mean that coal seams catch fire pretty much "as soon as" they reach the surface, in geological time.)
• More interestingly: how many coal-seam fires have happened historically in the Earth's 4.2 billion years being around, that finished and burned out all the coal, and now there's just a cave there? Or ex-cave sedimentary layer, given that these caves tend to collapse? Can we tell when such a cave/layer was previously a coal seam? (I assume so, because soot, but can pressure-treated soot be differentiated from other kinds of naturally-occurring mineral striae?)