Run the interface.
You cannot test a whole civilization. You can test its interface. Below is a toy of many coupled cells under shocks: feedback gets captured, stress spreads between neighbours, and the question is whether the Aeviternal mechanisms, decaying standing that unwinds capture and exit that cuts ties to captured neighbours, keep more cells inside their survivable band than a system without them. Flip the mechanisms off, inject a crisis, move the decay rate, and watch.
A toy, not a forecast. The cells, couplings, shock sizes, and rates are chosen for clarity, not fitted to data, so the honest finding is the shape of the behaviour and not the exact numbers: with the mechanisms the coupled system holds and recovers from a crisis, while without them capture compounds and spreads. The standing-decay rate has a tunable middle, since too slow lets capture compound and too fast leaks the standing that powers correction, the in-browser form of falsifier F–04. The full model, results, and limits are in the companion paper, The Partial Harness.