Born-level marginals can remain stable while trial-resolved measurement records exhibit order- and rate-dependent signatures. Such sequence sensitivity is invisible to histogram-only analyses yet constrains which operational descriptions of measurement remain viable under controlled schedule manipulations. Building on the Δt-sweep schedule witness introduced in a companion Foundations of Physics submission, this preprint poses a quantitative question: how much trial-resolved memory is minimally required to support the observed sequence sensitivity while remaining compatible with Born-stable marginals? We introduce (i) trial-resolved memory witnesses based on conditional mutual information (with a computational proxy), (ii) operational procedures for extracting conservative lower bounds (“floors”) on memory and occurrence using drift-robust null constructions, and (iii) an exclusion ladder that organizes control requirements for separating readout-only, latency/DAQ, drift/interleaving, and weak-memory explanations. The contribution is an operational lower-bound framework; it does not identify a unique physical mechanism and makes no ontological commitment. Device-specific attribution is deferred to experimental implementations. This manuscript is a working preprint subject to revision.
Emre Koçana (Tue,) studied this question.