[StuartBenjamin]

Problem. With sas_flag, sing_lim sets qlim from qmax and starts the Newton search for psilim at the grid point whose q is nearest qlim, anywhere in the plasma. When q peaks inside the plasma rather than at the edge (e.g. q ≈ 5.2 at the axis and 4.75 at the edge), qlim lies above every edge q and psilim converges next to the axis (psilim ≈ 2e-5). RDCON and STRIDE then solve on a sliver: the ODE integration fails and no rational surfaces are found.

Fix. Same change in dcon/sing.f, rdcon/sing.f and stride/sing.F: qlim steps down from min(q(psihigh), qhigh) instead of min(qmax, qhigh), and Newton starts in the outermost grid interval that holds qlim. When q peaks at the edge (qmax = q(psihigh)), qlim and the Newton start are unchanged, so normal equilibria give bit-identical results.

Test (vs develop)	Result
docs/examples DIIID and solovev resistive examples, RDCON + STRIDE, n = 1–4 (16 runs)	all outputs bit-identical
DIII-D equilibrium with edge-peaked q, reversed shear (176008 at 524 ms)	bit-identical (300 of 302 netcdf variables; the other two are cpu/wall time)
DIII-D equilibrium with q peaking at the axis (176275 at 1056 ms, two equilibria)	psilim 2.2e-5 → 0.989 at n = 2; n = 1–4 now complete