But rational expectations is just "people make error of random size and magnitude when predicting stuff". It's an absence of auxiliary assumptions about strategy, that's more compatible with complex emergent behaviour or people following a range of heuristics than the various forms of [implicit] consumption/production/etc function that preceded them. Ratex isn't "there's a single rational strategy", it's "these alternative specifications rely on a single fundamentally irrational strategy". You can still get path dependent outcomes with ratex models tending towards equilibria, just like many economists have managed to build many simplistic and wrong models of human behaviour with other specifications of expectation.

Anyway, epistemological agnosticism about outcomes is arguably much more compatible with the laissez-faire positions you seem to hate the most. If responses to and outcomes from public policy are unknowable, why do anything?

The classical physics analogy fits my position so perfectly I'm pretty sure I've used it before. Sure, Newton's theory of classical mechanics has no concept of subatomic particles and quarks simply don't behave like it suggests. But just because it's oversimplification at best and fundamentally wrong in other areas, and careless physics costs lives doesn't mean "caring about stuff falling down" and handwaving about complexity theory will build better bridges than the gravity-obsessed professional engineers you've confined to the cruise ship.