After extensive fooling around with a program along the lines of what I discussed earlier, I'm ready to conclude that a) the system's most fundamental flaw is the increasing number of failures needed based on complexity - it would make much more sense as a flat 4 - and b) you have to go pretty far with your assumptions of players aiding each other, choosing skills well, or getting DM's best friend bonuses to make the success rate reasonable for skill challenges with an average DC of 20 given to a 2nd level party.
Even if you let the players use a secondary skill 2 times out of 3 to give a cumulative +2 to the next primary skill check, had them always roll their best skill (sometimes as a secondary, sometimes as a primary - assuming that skill is +11), and gave them a 50% chance of a +4 bonus from things like racial benefits, utility powers, DMs best friend and any other situational thing you want to think of, the result is still a 30% failure rate for complexity 2 challenges (which is actually reasonable, I think), but only a 21% failure rate for complexity 5 challenges, which is a problem. Change it to a flat 4 failures and the numbers are 15% failure rate for Complexity 2 (still reasonable - the party would beat 2 monsters of their level more often, but failure has harsher consequences againt monsters), and 52% failure for Complexity 5 (which is harsh, but not insane, again success is a bonus and failure only a penalty, not a game loss).
If you use average DCs of 15 for that same party, you can give half the situational bonuses half as often, have the party make primary checks half the time (instead of 1/3) and let players use skills they are trained in but not necessarily amazing at (+8 instead of +11) 2 times out of 3, and you get 23% failure for Complexity 2 and 13% for Comp 5 (11% and 40% with flat 4 failures), which is probably about where things should be.