But where’s the positive evidence of what they’re [Austrians] claiming?

Says Krugman.
Right here, page 40, the ratio of debt to various measure of performance, charts on 6B.

These measures are all heteroskedastic, the variance expands as the economy moves forward.  If the economy really used use minimum variance then that should not happen. We do not measure expected variance into the future, we measure expected redundancy of items over a series of events. You might say the economy measures expected variation in events in the future.

The research referenced concerns the expanding debt crisis that middle class households sufferer during the slump.  Households measure that a series of paychecks may not happen in the future, so they prevent the car purchase event from happening now.

Why don't events happen smoothly?  Because events are quantized and as the negative event passes down the chain the quant sizes and/or inventory cycles we deal with are adjusted and no longer match the original, hence the new quants do not completely cover the distribution, it has to snake down the distribution.

Measuring redundancy when quant sizes change mean we lose resolution, we do not have the infinite dimensionality that expectations theory requires, so we end up with noticeable gaps.

I use the ice skating metaphor.  Imagine skaters in equilibrium, the faster skaters our on the perimeter, the slow ones in the center.  The velocities of the skaters is banded because skaters have to leave a finite amount of space to avoid collisions.  When a slow skater exits the center, he has to speed himself up in specific quants as he exits to the perimeter.  Doing that causes a queue bulge in each concentric velocity band. the queue bulge will appear as a spiral.

Let me repeat.  Keynes thinks we have smooth flow, so he uses minimum variance estimates.  Go back to Hilbert space, that assumption requires infinite dimensionality, it implies least squares estimates.  The economy cannot get 1.234 units of car, we cars in units of 1,2,3... Channel theory assumes minimum redundancy of events, we count in integers, and even then the integers are sparse.