# R-01B protocol, before new model computations 2026-11-14. Preserve R-01A outputs and raw inputs. No UI, new clone acquisition, or 100-clone ensemble. Objective is replication/sensitivity, not a forced seven days or a preselected award story. ## dGR203 controlled ladder Primary control: R-01A published rounded coordinate, nominal 5 m SL2 + fixed 80/240 m far field, 2 m height, explicit DE421 ME frame and DE421 trajectories, geometric centers, horizon 250 km, event solving on 300 s discovery brackets. Literature-date representation follows R-01A: [2026-11-06 00:00 UTC, 2026-11-28 00:00 UTC). Clipped durations end at the right boundary. A boundary does not represent a measured terrain transition. Record censoring. Before outcomes, register one-factor comparisons: - Temporal: 12 h, 1 h, 10 min discrete left-endpoint cells, plus refined events. Discrete cell durations sum true cell widths clipped to the domain; they are not proof of uninterrupted visibility between samples. Also report the span between first and last true samples as an aggregation diagnostic. Do not attribute either convention to QuickMap without evidence. - Endpoint sensitivity: end Nov 27 00:00, Nov 27 12:00, Nov 28 00:00 UTC. These bracket date-label implementations; only the inclusive-calendar-day version spans 22 full days. Paper Table 2 reports 22 days but exact timestamps are unknown. Integer floor/nearest rounding are diagnostics, never fitted truth. - Terrain: average nominal SL2 to ceil(20000/60)=334 square cells, preserving exact extent; actual resolution 59.88024 m. Keep the far field unchanged. This is a reproducible local resolution proxy, not the undocumented QuickMap source DEM/pyramid. Evaluate Product104 separately at the same location. - Horizon cutoff: 150/200/250 km at otherwise fixed nominal assumptions. - Height: 1/2/5 m, with 2 m the replication baseline. - Visibility: geometric center, finite spherical-disc any-visible, and full-visible. Account for the entire azimuth-dependent skyline, not merely center elevation +/- radius. Radii must come from an authoritative source; Earth sphere approximation and absence of illuminated-phase weighting must be explicit. No equivalence to QuickMap is implied. - Coordinates: a separately labeled diagnostic at the original archived point instead of the published rounded point. This is a position control, not a terrain-model factor or a silent replacement of dGR203. - Ephemeris: assess authoritative DE441 frame/translation compatibility and resource requirements. Do not load a mismatched orientation or substitute DE440 under a DE441 label. A documented not-run blocker is acceptable. For each run report Sun, Earth, simultaneous longest duration and start/end, censoring, differences from the same-domain R-01A control and seven-day reference, and >=6.5 and >=10 day outcomes for each metric. The 10-day paper screen is primarily solar illumination; the same threshold on simultaneous visibility is a separately labeled diagnostic. Numerical checks: reproduce R-01A same-domain center to 0.1 s; finite-disc flat-horizon analytic case, zero-radius limit, skyline extrema and any >= center >= full interval inclusion; compare finite-disc root discovery at 300/150 s and spatial skyline integration as needed. Material numerical instability blocks scientific interpretation. No causal decomposition of interacting factors from a one-factor ladder. Named joint replication proxies may be added using the already registered settings and must be clearly joint. Verdicts: A only if independently documented implementation choices and archived results substantially reconcile the discrepancy; B if reproducible effects explain some gap but material/semantic uncertainty remains; C if no responsible explanation is supported. A numerical seven-day match alone is not evidence of the paper's undocumented implementation. ## Multi-site rule, to execute only after the dGR203 audit Use the original DOI-linked Wueller candidate archive, 2026 published simultaneous duration and supplied point geometry. Resource/coverage condition: inside existing SL2 and A304 valid footprints, with the existing far field. This intentionally restricts inference to available high-resolution coverage. Select up to two NEW sites per published-duration band [6,7] days (near 6.5) and [9,11] days (near 10), ranked by absolute distance from the relevant threshold then lexicographic site name. Exclude dGR203 from new selection but retain it as a control. Empty slots stay empty; do not widen bands after seeing outputs. Record all eligible/excluded sites and coverage. Published durations choose sites, never uncertainty outcomes. Use archived exact point coordinates for new sites and record that difference from the dGR203 control. Run nominal comparisons only: existing 5 m/nested center event solution, approximately 60 m local proxy center, and 5 m finite-disc any-visible; all at 2 m and 250 km in the same 22-day control domain. No new ensembles. Recommend a later five-clone pilot only if a site crosses 6.5/10 across defensible configurations, or lies within 0.5 day of a threshold, after numerical checks. Any prioritization uses model outcomes only after the frozen site list exists. Taxonomy is provisional: separate model-form/configuration, data-representation, and operational axes; unsupported axes remain unassessed. No weighted score. ## Post-screen mechanistic control (registered after detecting dGR101 contrast) The metadata-selected dGR101 has a large 5 m/coarsened difference, so verify that result numerically before interpretation. Then perform one transparent post-screen control: keep the observer's absolute radial elevation fixed at the nominal value while replacing terrain with the coarsened representation. Compute the required height above the coarse surface algebraically, not by fitting a duration. Run only if that height is positive. This decomposes observer-datum shift from changes in nearby terrain shape and is not a new preselected candidate or a claim of exact QuickMap behavior. Preserve both the original 2 m comparisons and this separately labeled diagnostic. Extract bracketing horizon witnesses and existing LDEC/uncertainty samples for attribution; do not infer measurement benefit or success probability. The fixed-absolute observer diagnostic unexpectedly moves the coarse result below 10 days. Preserve that outcome. Validate its solar/simultaneous conclusions with 150/75-second timing and joint halving of azimuth/radial sampling, using the same 600-second spatial tolerance; retain Earth-only checks separately and do not interpret unaccepted Earth-only timing. This is post-screen validation, not a new selection rule or a fitted model parameter.