Can you see the changes?
After some help from my wife, I managed to spot this also in a running FG instance...
It looks for me that the shader calculate something in reverse distance
Problem is, the lights go all via the same shader, so if an attenuation is with one of them, it is with all of them, I can for instance instruct the shader to draw all lights 10 times larger, and I get this:
Relative size of lights is pretty much unaffected and so is the fact that they get smaller with distance. Note that I do not see the blinking yellow lines even here.
I can also instruct the shader to draw all lights at a constant size of 20, and interestigly enough I get this looking at the missing lights
Now I do see the line blinking like this, but it is not actually being processed by the fragment shader, unlike the other lights which have a tight core and a faint halo here the whole square of the point sprite is drawn in a homogeneous color.
So it would seem the problem is with the input parameters - when not instructed to draw at constant size, the vertex shader doesn't really know what square to make - max-size, size and min-size don't seem to be set, which explains the weird distance behavior - the code has invalid parameters to compute size.
There might be more parameters missing because obviously the fragment shader also does not know what to do with the sprite, so it appears as a homogeneous square even when drawn at constant size.
(Edit: Actually there's not more missing - with the given input, the fragment shader believes that we're smaller than a single pixel and hence colors the whole surface in uniform color - if I disable that size check, the light is drawn fine - so it is really only the input size definitions which do not appear as they're needed).
My conclusion is that this is not an issue with the GLSL code but with the C++ part which sets the parameters for these effects (unlike in many other cases, here size, min-size, max-size and attenuation are not filled by the effect file but directly set by the C++ code), and something seems to go wrong for these lights.