>> Will geometry kernels ever be fully robust, or is there still a holy grail to be discovered here?
WRT robustness maybe not. Even the commercial ones tend to fail in certain situations. I'm not sure if there's a widely desired holy grail people are looking for and I don't even know what that would look like. I might recognize it if it appeared, but maybe not.
BTW for FEA I think a major new thing is available in Altair products the last couple years.
What about one of those SDF code CADs but with FreeCAD's sketcher as the UI?
I'd imagine extruding a sketch into an SDF object is possible, and you could just only allow sketches to reference stuff in other sketches, or the bounding box cube of a sketch extrusion.
They'd need some better tools for referencing stuff between sketches but other than that, it seems like it's mostly all there.
And then at the end if you wanted a STEP I'm sure someone out there has got to be working on some kind of mesh to STEP thing that properly preserves design intent.
I am pretty sure that this is a consequence of the mathematics for modeling this being _very_ hard, and a robust (and performant) solution requiring as-yet undiscovered/unpublished mathematics.
WRT robustness maybe not. Even the commercial ones tend to fail in certain situations. I'm not sure if there's a widely desired holy grail people are looking for and I don't even know what that would look like. I might recognize it if it appeared, but maybe not.
BTW for FEA I think a major new thing is available in Altair products the last couple years.