I read an article in Scientific American about the natural fission reactor in Oklo, Gabon. Most of all uranium is U238, and today we find 0.72% U235, the isotope useful for chain-reactions.
U235 has a shorter halflife than U238, so in the past the ratio was different. 2 billions years ago it was about 3% U235 which is the concentration we feed into reactors today. Back then (2 billion years ago) a reactor could form naturally if there were enough uranium in one spot and some water to slow down the neutrons.
Going further back in time, there should have been about 25% U235 when earth was formed 4.6 billion years ago. Weapons-grade uranium is 20% or more. Could there have been spontaneous fission-explosions in earth's early history?