@aparrish this is awesome stuff! I do optimization usability research, and can commiserate: jury-rigged hybrid algorithms perform annoyingly well in engineering too, unless there's a clear convex subproblem (...falling into the shop talk singularity...)