However I don’t think this issue would actually matter, because from the observer’s point of view in such a situation the probability of discovering you are in one of the Everett branches where the probability-1 event did not occur is exactly 0, so in terms of forming beliefs, you’d assign zero probability to that outcome.
Think of it this way, if I create a random continuous function on [0, 1], the probability that it is also differentiable is 0, even though there are tons of familiar examples like x^2 or sin(x).
So if I offer you a wager of whether the randomly generated function is differentiable, your belief ought to be that it always is not, even though the technical term would be “almost always” in formal measure theoretic probability theory.
I haven't finished reading the paper, but does anyone who has already encountered it know if this objection is covered (either in the paper itself or in later discussion)?