This is a known issue since day one and one I’d like to fix. Having a tube light oriented along the spot axis is basically useless. I’ve had very little time to work on area lights in general unfortunately which is why it is the way that it is. I only had the time to implement the omni light version correctly, not the spot version.
IES are a more complicated matter. Technically the measured data stored in the IES is coming from a point. It is measured in polar coordinates from that point. It is ambiguous how to use that data when the source shape isn’t a point. Is it angle from center point? Is the IES profile integrated over the area? Is it applied from the closest point on the surface of the shape?
“So, in general, for a spotlight I would not recommend adjusting the Source Radius or Length if you are trying to duplicate real world conditions.”
I disagree with that. We simulate spot shapes as shadowing functions as if the fixture is blocking the light outside of the cone angle. This is different from a reflector or lens focusing the light. If we did the later the center of the spot light would get bright if the cone angle was smaller. With this in mind having a sphere shape source is totally valid. A spot light probably makes more sense to be a disk but again I haven’t had time to implement that. If/when I fix the length axis to tangent to the spot axis a tube light would make sense as well. I’ve worked on rect lights in my free time but they are WAY harder. No one at this time has a good solution for them.