The first images shows an internal issue with dynamo/revit, I think re-installing a stable version would solve it (I had faced it a while back when I wanted to work with geometry related nodes, and only re-installing dynamo fixed it).
Were you running Dynamo 2.0.3 with Revit 2019 or Revit 2019.1? It’s a known issue that Dynamo is quite buggy with Revit 2019. I especially remember toubles with the Python node, but I don’t remember if it also had troubles with the geometry engine.
I’d amend that to say "it has some issues with early ships of Revit 2019. It does quite well with 1.3.4 which I believe is what it ships with, but it’s been like 18 months since I installed it so I’m not sure offhand. 2019.1 is now 13 months old, and 2019.0 is 16 months old.
For now I would make sure that everything is up to date, which at this moment means Dynamo 2.0.3 and Revit 2019.2.2.
FWIW, I am not a big fan of this method. It either fails to catch issues (ie: a pipe running parallel to the edge of the tread, or a wall mounted light fixture that is at the wrong height, a floor that extends over the first 1/3 of a riser), or requires too many points to catch everything you’d need (ie: a 1/8"point grid at each walking surface.
Instead I recommend looking at the geometry of the stair to pull the walkable surfaces (area >= stair width * tread depth && normal at parameter 0.5,0.5 has a Z component of 1) and pull the leading edge of each to build the actual clearance shape, and move that into the rvt file with associated element bindings so that you can update them. Once that mass is added to the file you can use the interference checker in native revit, or any of the many other clash detection tools (navisworks, glue, model coordination, dynamo’s geometry conversion and geometry.does intersect, bimorph nodes, etc.) as you would any other element set.
Thanks for your comments on Ray bounce method. I agree.
In this case, I just wanted a quickndirty solution to use for one project that had a tight deadline and just needed an easy overview solution.
The methodology that you outline sounds very robust. I will try that when I get time.