Rotation into the Portfolio Trap: How Stakes Order Basin Fragility

A recurring account of persistent control deficiencies is strategic inattention: a rational monitor withdraws observation from the controls that matter most, because looking concentrates blame, and does so most where the stakes are highest. That blind spot has been characterised statically. This paper puts it on a time axis. On a productive substrate, where observation is valuable in its own right and each control carries an engaged basin separated from an inattention trap, we show that—holding the responsibility regime common across controls—the highest-stakes control has the thinnest engaged basin: it is the most fragile. Rotation, a recurring downward shock to a monitor's observability capital, therefore drives the costliest controls into the trap first, and, by the metastability of the engaged state, durably. The policy is correspondingly sharp. Escape is a basin crossing, not a marginal repair: a safe harbour must narrow the de facto responsibility differential past the spinodal, not merely past the Maxwell point at which engagement becomes worthwhile—a half-measure leaves the highest-stakes controls blind. The clean stakes ordering is a homogeneous-responsibility result; off it, fragility is governed by a two-variable index $\lambda_b(\psi, \Delta)$ and can invert. The stochastic counterpart of the result is deferred to a designated companion.