It may indeed be impossible to be a believer if you have very high levels of epistemic rationality, but it's compatible with very high levels of instrumental rationality combined with moderately high levels of epistemic rationality.

Agree. (Or at least I agree about the instrumental rationality compatibility. The "moderately high levels of epistemic rationality" would depend on the design of the metric.)

Edit: Other replies reminded me I may have been hasty in my agreement. Perhaps put the instrumental rationality compatible in with "depending on the standard of measurement". Simply because that belief puts some hard limits on how instrumentally rational the individual can be. Unless the belief is so compartmentalised that they do things like still actively work to combat existential risk, at least as it applies to themselves or otherwise act as if they are taking such concerns into account in their decision making. I maintain my endorsement with your general sentiment.