It doesn't, and it can't. A SL of 2 represents a certain ability, and if a player improves beyond that ability they become a SL3. A SL of 9 also represents a certain ability, and if a player improves they stay a SL9, meaning they should be expected to win more often than before they improved against someone with the ability of a SL2. Taking into account that SL2->SL3 movement happens much more frequently than SL8->SL9 movement, that means SL9 will always have an advantage over the other skill levels. SL1's and SL9's are special that way, but SL2 through SL8 is pretty much as "even" as you can get. Not at any point in time for specific players (that depends on whether the two players are "trending" up or not, because SL reflects measured ability and that means you have to demonstrate the ability of a higher SL before the system makes you one), but over the long run for random players in each SL range it's very close.
This is the goal, though I would substitute "measured ability" for "average". At the low skill levels, it is easier to do that than at the upper skill levels (the upper limit is higher for lower SL's, relatively, because the measurement is so low), but as one improves their consistency should improve too, making the lower limit higher. In other words, when both players are "on" the lower skill level has a slight advantage, but when both players are "off" that advantage flips to the higher skill level.
You can say that about any handicapping system. Sandbagging means not showing your true speed until you need it. As the stakes get higher, you're more apt to show it. To catch the sandbaggers locally, if they never show their true speed, means you have to be willing to be wrong sometimes and punish innocent players. That's worse than letting a guilty player get away with it, if you don't know they're guilty. It's also why most of us consider data from outside our own system (like performance in tournaments and other leagues, or Fargo ratings) to help with the subjective part of our jobs, because catching a good sandbagger is necessarily subjective.
None of this is a trade secret, it's just common sense...