Das Morbo,
I am not sure, if the LOS-tool considers ground level only. Because then tanks couldn't be placed well in hulldown positions: if the LOS-tool would show hulldown, the tank would be 2 or 3 meters higher and completely exposed. I don't remember to have such extreme hulldown positioning problems in the open - my impression so far has been, what the LOS-tool promises, roughly is delivered.
I believe that you, das Morbo, made a very interesting point, why there can be such catastrophic, game deciding, problems with the LOS-tool with trees involved: if not the whole vehicle is taken into account for the LOS-tool, but only one or two spots?
Since the LOS-tool can differentiate between the blue and grey line, there must be a mechanism that uses more than one spot for the LOS check. A lower one and a higher one point of view. That the tool can differentiate between full and partial LOS seems to be an indication, why the LOS-tool is NOT only calculated from the ground level or from a single spot. Otherwise it couldn't differentiate between fully and partial LOS.
But how could the behaviour above - with trees involved - be explained, if we exclude a bug?
The Pz. IV is close to trees. For the spotter being closer to an obstacle, the obstacle blocks a wider angle of vision - just like if we place ourself behind a tree. Depending how close we are to the tree, it's entirely possible that we don't see anything in front of us. Even if climbing on a ladder at the same spot, wouldn't change LOS.
So if the LOS-tool would check one or two other hight levels above ground but from exactly the same position, then the tree would still be in front and the result would be the same: tree blocking LOS -> secure.
But a vehicle or a tank is not only a spot, but has a big size. So while the LOS-tool shows LOS blocked, it could be possible - IF the LOS-tool would work that faulty way - that the tank could almost be fully exposed, while the LOS-tool suggests the user that LOS to the vehicle will be blocked...