There are inherent difficulties in recording digitally (sampling) as it causes discrete bits of music to be recorded in series (thousands of individual sound-bits per second). Because there is inherently a limit in sample rate and bit depth during computer recording, this method is likely to result in small errors.
And, more errors happen when you do any processing on what was recorded in (primarily due to multiplication, I always hated multiplication). If there are enough of them (and there are many quiet samples), it can make you audio less representative of what you recorded in (no good).
Adding a very small amount of (relatively) continuous noise helps to (kind of) smooth these errors and allows for better dynamic range in a track.
Whether it is a good idea or not is probably not important until everything else is prettied up. Really, that is the bit that I have not gotten figured yet as I am much newer to recording than to the tech that is used.
Anyway, for anyone who stayed awake during this, I hope it helps.
Edited for clarity.

