Hi Steve, that is an odd result. I believe the problem basically stems from trying to make a tube that goes around a bend of the exact same radius.

That basically results in a surface that intersects itself, just kind of grazing itself right in this one point here:



Various mechanisms can behave unpredictably in the presence of self intersecting surfaces like that, so it's something that you generally want to avoid - if you use just a fraction of a smaller radius than those bends so there is at least a tiny clearance in those corner areas you will likely get a lot more predictable behavior.


Right now the surface is so close to bunching up and overlapping on itself in that area, and in some cases it seems to end up crossing over itself just a bit more than in other spots, that's what would lead to some slightly inconsistent looking behavior with some area happening to work ok and others maybe not.

All floating point mathematics that are done on the computer are subject to slight rounding errors so when you have a situation where it's just barely ok just the slightest difference in the sequence of numeric operations can possibly produce a result on one side of the fence or the other side.

- Michael