Record, Erasure, and Distributed Witness Sensitivity on IBM Quantum Hardware: A DAGI Validation Study

This manuscript reframes a series of IBM superconducting-qubit experiments as a targeted validation study for the Directed Acyclic Graph Interpretation (DAGI), with emphasis on the operational distinction between \emph{record creation} and \emph{record erasure}, and on how that distinction manifests in different classes of observables. DAGI models quantum processes as a directed acyclic graph of informational events: branching is permitted only when an interaction produces a durable record at the relevant coarse-graining scale, while coherent erasure can restore interference by removing the distinguishing record. We consolidate three falsifier-friendly hypotheses: (H1) remote record activity need not produce a universal ``clock slowdown'' for arbitrary local probes; (H2) remote influence is expected to appear preferentially in observables supported by a relational/entangled substrate; and (H3) the record/eraser distinction should be directly visible in distributed witnesses under strictly pulse-matched controls. Using the C1 ``informational time dilation'' series as the local benchmark, and the T4 remote-closure suite as the nonlocal stress test, we report a split outcome: clock-side remote slowdown remains bounded and proxy-dominated in tested geometries, yet a GHZ-class distributed coherence witness exhibits a large, replicated record-creation (REC) versus reversible sham (SHAM) separation on both ibm_fez and ibm_marrakesh, and is strongly restored by coherent erasure (ERASE). We position this pattern as evidence for observable-selective remote sensitivity that is consistent with DAGI's record/eraser mechanism while remaining conservative about hardware backaction channels.