code, frozen artefacts and per-seed data for reversible power probing in Open RAN

Code and data accompanying the article "Probe Before Repair? Reversible Power Probing for Fault Diagnosis in Open RAN: A Preregistered Simulation Study". An ns-3 (ns-O-RAN-flexric, ns-3.42) scenario contrasts two faults that degrade the same users: a loss of serving-cell transmit power (F1) and a fixed-MCS link-adaptation misconfiguration (F4). A short, reversible transmit-power probe is tested for decision value against a passive comparator on the same random streams. All 630 runs passed a five-level quality-control chain, and every decision rule was frozen before the streams that tested it were generated. Two preregistered confirmatory tests are included. C2 (+3 dB probe, 40 streams): the probe response was fault-specific and cheap but did not improve decision accuracy (59/80 vs 57/80 correct; exact McNemar p = 0.75). C3 (+6 dB probe, 40 streams screened to be healthy before the fault): accuracy rose from 56/80 to 70/80 (discordant pairs 14 vs 0; p = 0.0001), but a third user was delayed beyond the preregistered 10 ms margin under F4. Contents: pinned simulator source; quality-control, feature and analysis code; calibration plans; the frozen model contract and decision rules; both preregistrations with their amendments and SHA-256 freeze lists; per-seed summaries and aggregates for all runs; the original preregistration (C1, v0.7), withdrawn before any run, together with the finding that led to its withdrawal; and all figures with the scripts that generate them. Raw packet traces are available from the author on reques.