Global optimality or approximation guarantees for the general-regime policy

Establish global optimality or an approximation-ratio guarantee for the AquaMend general-regime policy that jointly handles correlated failures, shared recovery operations, and partially resolving probes.

Background

AquaMend’s general-regime policy evaluates one re-probe followed by a terminal recovery decision using a joint posterior and expected loss. Unlike the structured regime, it does not optimize a complete adaptive policy tree, and the presence of correlated failures, shared recovery operations, and partially resolving probes prevents the per-belief decomposition used to prove the closed-form result.

The paper explicitly states that neither global optimality nor an approximation-ratio guarantee has been established for this policy. The unresolved problem is therefore to provide either an exact optimality result or a provable approximation guarantee for the general recovery setting.