A finite-separation binding well in a frozen sextic Q-ball model: independent-code confirmation in flat space, first full-GR evolution, and a boundary-drainage triage

I study two-soliton interactions in a frozen complex-scalar model with a sextic self-interaction (parameters m = 1, λ = 1, γ = 0.5 throughout), using the ω = 0.85 Q-ball (f(0) = 1.0633, E = 337.64, Q = 365.45, E/Q = 0.9239) as the reference soliton. Earlier flat-space measurements reported an interaction-energy minimum — a binding dip — for an at-rest charge-conjugate pair at separation d ≈ 7–7.5, ≈21 units below the two-free-soliton threshold. I confirm that dip in an independent production code (a BSSN evolution lineage run at G = 0) against four criteria registered before the runs: the same-code static curve has its minimum at d = 7 with depth −21.17 (threshold 673.581), and the masked t = 100 evolution energies preserve the required ordering, E_m(7.5) = 563.943 < E_m(6) = 571.395 < E_m(14) = 592.166 against a masked threshold of 645.242. Evolutions at all three separations end as bounded, oscillating two-peak structures; the pairs shed energy radiatively through the open boundary and are trapped by their own losses. In full general relativity (G = 0.001) the well's shape is not reproduced at build level (E(7.5) − E(14) = +0.24 against a pre-stated ≤ −3 bar), while a t = 100 evolution of the d = 7.5 pair remains a bounded two-peak structure with modest losses (E: −3.272%). A registered settled-vs-settled comparison, the designated judge of the dip's GR status, has been run and returned UNCLEAR: no core separation reached a settled state within T = 200, so no settled comparison exists (§5.3). A second registered phenomenon — a gravity-induced flip of the anti-phase v = 0.2 collision outcome — could not be judged: the production code's G = 0 control binds the pair where the map code passes it through. A two-step triage exonerated the initial-data construction and identified the open boundary's radiative drainage (12.467% of whole-box energy shed, vs 0.069% in the map code's reflecting box) as the prime suspect; the decisive boundary test has since been completed (§6.4): the larger-box arm shed less (7.223% — the shed is strongly boundary-linked), but its pair drifts slowly outward rather than binding (final-third slope fit +0.0623; plain UNCLEAR, supplementary ESCAPE-TRENDING), so the boundary is not exonerated as the decisive ingredient on this record, the map-versus-production split remains open between boundary, scheme, and timestep, and the flip cells remain held. All criteria in this programme were recorded before the runs they judge, and all failures are reported alongside the passes.