I like bagchucker's suggestion above - that the salaries be renegotiated based on historical SIM success, not popularity. That way the cookies that always seem to perform well for whatever reason will see their salaries rise. One of the biggest complaints I see in my leagues' message boards are players severely underperforming - the foremost reason I see people losing interest in their teams. There are a lot of reasons for that, of course, but the expectation is that if you spend $12MM on a pitcher he should perform better than a $7MM pitcher, despite the various factors that may arise such as home field, strength of division, etc. But they too often don't, which in part nullifies owners' skill in creating a team under a salary cap. Yes, it's a lot of fun when a player overplays for you, but if I spend $8.5MM on '75 Joe Morgan, I expect that he'll hit better than .250.
Relief pitchers are a great example. I've found that it really doesn't matter if I spend $4MM or $2.5MM on the relief pitcher because it's a craps shoot as to how they'll perform. When, truthfully, since $4MM is at the top of the range for RPs, and assigned generally to the elite ones, they should perform much more reliably. By adjusting salaries based on SIM performance, while at the same time adjusting the salaries of those who generally overperform, you'll be getting a much more true picture of players all across the league at all positions.
If the salaries were based on SIM performance, and updated every few months - that makes sense. Then, you're actually increasing the challenge of drafting strategy, because when owners gravitate toward cookies, they'll be paying more for them, forcing the decision of whether to continue to use them and weaken the rest of their team, or change focus and use salary cap room to strengthen the rest of their team (thus strengthening the rest of the pool by making cookies less attractive). The goal of the salary cap is to challenge owners to build the best team with the money available. The luck of the draw is important when it comes to the actual simulations, but I'd like to see the results lean more heavily toward drafting skill than luck. I think this would help.
I don't mind a new approach, though for my leagues, my biggest concern with the proposal as it stands is changing salaries once a league's already started. I think there would have to be a way to lock in a salary structure for a league once it's established.