Here's some good technical stuff on these types of sensors I found...
This causes the device to measure the effective "center of gravity" of the light hitting the sensor. So the type of effect I was talking about could happen, but it's a function of how small the focused dot is on the surface of the sensor. The smaller the dot, the less effect there would be.
The first technical document above goes into great detail in how to design such a sensor including all the math for selecting lenses and calculating dot sizes. I read a lot of it but not all of it.
Laser diodes it seems from the document do have extremely small light source sizes (20x4 um) which can make for an extremely small dot on the sensor if the optics are designed correctly and the dot is focused accurately. However, as the more the target moves, the dot goes more and more out of focus. So the dot size will be a function of the design, and how much the target is allowed to move. So if you want the sensor to only detect a range of movement (like the thickness of a sheet of paper) from a distance of 10 mm then the optics can be precision focused to make a very small dot on the sensor surface and create a highly accurate reading (a .004 cm dot on a 3 mm sensor surface just doesn't cause much error when half the dot goes dark and shifts the center of gravity of the dot by .001 cm).
But if you are talking about allowing the target to move 100 cm and you aren't using some complex movable focus lens system (which I don't get the impression these devices use), it seems to me it's not going to be able to focus the dot very accurately on the sensor for the full range and it could create a very large dot on the surface and the PSD will then be calculating the center of gravity of the dot. So if the out of focus dot expands to 2 cm on a 3cm sensor, and half the dot goes dark as it passes over a edge on the target, it could cause the center of gravity of the dot to shift by .5 cm which would be something like a 16% error.
So, end result, as I suspected, there will have to be some error caused by a sharp reflective change on the surface just based on how these systems work. But without knowing the exact design details of the sensor, I couldn't begin to estimate whether the effect is a .001% error or a 20% error.
Here's another idea I just thought of for caring about the direction of movement. If there is a sudden change in the depth of the target as it moves (you scan past a step in the surface), then it's possible that the dot could be momentarily hidden from the view of the receiving sensor (which is at a slight angle from the transmitter). This could cause a slight blip in the distance reading. Mounting it as they suggest could prevent the blip. Maybe both this effect and the transition from light to dark surface just causes more "blips" in the reading unless you mount it as they suggest relative to the motion. And if you had a stair-step effect in the right direction combined with a color change, you might get double the effect. (like scanning a light object going past on a dark conveyor belt).
For applications where the target is well known, I can see where effects like this might make it useful to mount the sensor in a way to minimize the noise caused by these effects (like measuring the thickness of boards going past on a conveyor belt). But for robotics use where you can't control the environment there's going to be so much noise and unpredictable results that the minor effects that I talked about seem like they would be insignificant making it of no real concern which way you mount it.
How much do these sensors cost? Where can you get one cheaply? I'm feeling motivated to buy one and play with it now....