Firstly may I point out that the readback process simply isn't particularly sensitive to small amounts of stray light.
It depends on interferometric cancellation of light in the pits and lands.

Now taking this into account and understanding the Reed/Soloman algorithims you will find that if a disc is in a reasonable state of repair the system will read the bits off of the disc with an error rate of approx 4-5 UNCORRECTED errors for every 10 Million bits of data read. That equates to one ever 3-4 mins on average.
In reality that is perfect data reading.

So how does a surface treatment improve on virtual perfection?

I dont know if you are interested or have acces to this type of material but here are a couple of good references on CD playback.

Blesser, B., "Digitization of Audio: A comprehensive Examination of Thoery, Implementaiton and Current Practice,"

J. Audio Eng. Soc., vol 26, no 10, 1978
Oct.


Vanderkooy, J. and Lipshitz, S.,
"Resolution Below the Least SIgnificant Bit in DIgital Audio Systems with Dither,"

J. Audio Eng Soc., vol 32, no. 3, 1984 March.

One final one which is my favourite is:
Principles of Digital Audio, Ch. 9,
"The Compact Disc: Player Optical Design,"
McGraw Hill 1995

Stuart Stebbings