Beliefs which cannot be updated aren't useful, but not all beliefs which might reasonably form a "worldview" are un-Bayesian. Maybe a better way to talk about worldviews is to think about beliefs which are highly depended upon; beliefs which, if they were updated, would also cause huge re-updates of lots of beliefs farther down the dependency graph.

Yes.

Beliefs have hierarchy, and some are more top-level than others. One of the most top-level beliefs being:

1. a vast superintelligence exists
2. it has created/effected/influenced our history

If you give high weight to 1, then 2 follows and is strengthened, and this naturally guides your search for explanations for mysteries. A top-level belief sends down a massive cascade of priors that can effect how you interpret everything else.

If you hold the negation of 1 and or 2 as top-level beliefs then you look for natural explanations for everything. Arguably the negation of 'goddidit' as a top-level belief was a major boon to science because it tends to align with ockham's razor.

But at the end of the day it's not inherently irrational to hold these top-level beliefs. Francis Crick for instance looked at the origin of life problem and decided an unnatural explanation involving a superintelligence (alien) was actually a better fit.

A worldview comes into play when one jumps to #3 with Miller-Urey because it fits with one's top-level priors. Our brain is built around hierarchical induction, so we always have top-level biases. This isn't really an inherent weakness as there probably is no better (more efficient) way to do it. But it is still something to be aware of.