From occasionally repairing modern bicycles the way the chain is falling off "regularly" doesn't click with me: If one spoke is bent so much (more like a serious wheel wobble) that - at some point - it touches a "damaged" chain link; this is quite a spectacular failure.
I have to assume back then those bicycles were built quite differently that something like that could happen in such a predicatble fashion. Or along the way some information was lost and something different added in that anectode.
Curious to see other versions I have quickly traced back the ancetode to a Nature article "Are mathematician logical" (1987) by Ian Stewart[0]; the version/wording unfortunately here is quite the same.
Bikes back then weren't much different. Some from that time are still around and are not technically anything special.
I don't buy the story. It might be what he told himself, but what fixed it was just the maintenance in general. Probs moved the wheel and tightened the chain. In any case if the bump of a crocked wheel could throw his chain, it was waaaay to loose.
Curious to see other versions I have quickly traced back the ancetode to a Nature article "Are mathematician logical" (1987) by Ian Stewart[0]; the version/wording unfortunately here is quite the same.
[0]https://doi.org/10.1038/325386a0 (via paywall --> sci-hub)