I see there is no planarization in your definition, and the result does not have faces close to planar, so it doesn’t make sense to connect it to the face-face offset component, which requires an input mesh which is both planar and conical.
About the speed- that definition should run very fast - I suspect maybe you had the downstream part of your definition activated, which includes a Boundary Surfaces component on the offset mesh and beams. That would mean that for every edge beam and panel it is creating a mesh, then converting each one of these into a trimmed NURBS surface at every iteration.
More generally though, this approach of draping a regular grid over a freeform solid is not usually a great fit for creating planar conical meshes.
This sort of draping is particularly suitable for generating geometry for something like an active bending timber or composite gridshell with long continuous laths that are assembled as a flat regular grid then pushed up into shape or lowered onto supports (Weald and Downland museum and Mannheim multihalle are good examples). There a key thing is having equal edge lengths (a Chebyshev net), but the faces are usually far from planar and the vertices are far from conical. The outer skin for these types of gridshells is often something like a flexible membrane.
For something like a steel and glass roof structure, with each edge a separate member, bolted or welded at the nodes, and each face as a glass panel, a different approach is usually needed. To get planar faces and torsion free nodes, the grid needs to be very carefully designed to fit the shape (including the boundaries if you don’t want triangles) and its curvatures - more like a tightly tailored suit than a draped poncho.
Don’t be too discouraged if it seems difficult though - this is certainly not a beginner exercise.
There have been several discussions on here about freeform torsion free planar quad gridshells, but it’s easy to lose sight of the fact that this is still a relatively new topic of research and development.
Until only a few years ago, the only option to even attempt something like this would have been one or two specialist geometry consultancies.
I believe there have still not been any large scale true freeform conical PQ mesh gridshells actually built. I linked a few examples that come close here, but none of those meet all those criteria.
With the tools now in Kangaroo it is possible to create these torsion free planar quad meshes (such as this example I posted in that same thread), but doing so from scratch for different boundaries still requires a fairly good understanding of principal curvatures and mesh topology.