To me, the most underrated counterargument to the Pascal's Wager is that it ignores that your actions can have consequences on either side of zero. Once you take that into account, many things move from "seeming no-brainer" to "totally undecidable omg what the heck do I do"

(And yes, like all Pascal's Wager discourse in the 2020s, this has an AI safety tie-in. Soooo many actions done in the name of AI safety so far seem to have had very ambiguous consequences (Premise: all the other AI labs are reckless, I'll make one that focuses on safety! 2 years later: we have n+1 reckless AI labs))

n+1 reckless AI safety labs inevitably reminds of this relevant XKCD