I think you are engaging in the gambler’s fallacy fallacy. The well known gambler’s fallacy is to believe that past events influence later (supposedly) independent events. A run of “rouge” on a fair roulette wheel does not indicate that “noir” is “due”. But a long sequence of “rouge” does call into question whether or not the wheel is actually fair. Bayes would understand.
If you are waiting for a bus on a scheduled route, as each minute passes, it would seem increasingly likely that one would come in the next minute as the scheduling is designed to spread the arrival of buses a more or less uniform intervals, and that assumption is reason. It is based on knowledge of the state of system.
One the other hand, if one is counting decay products of a mass of radioactive material with a Geiger counter, one expects “statistically” (more accurately: “probabilistically”) a certain number of hits per unit time, on average. But the occurrence of hits is completely random, unlike the bus, one can have several hits in one five minute interval, and none in another five minute interval. The count in one interval has no effect on the count in next (assuming the observation interval is a small fraction of the half life).
The gambler’s fallacy is akin to likening the occurrence of “rouge” or “noir” (or heads or tails) to the arrival of scheduled events, when they are (given fair wheels or coins) independent, like the decay of a radioactive isotope. The gambler assumes knowledge of the state of the system which does not exist. I think you are making the same assumption. There may well be asteroids out there on predictable trajectories will hit in 100 or 1000 or a million years. But we lack any knowledge of their current state, or even about their existence.