The persistent disconnect over opt-outs seems to be around the distinction between how much additional potential value it adds at the time of signing for the player (which is substantial), vs. the probability that it results in an optimal outcome for the team (which is certainly non-zero). Everyone seems to agree on the possible different outcomes, but we aren't laying them out:
1.) player does well, opts out, signs with other team does well [winner: player] (JD Drew)
2.) player does well, opts out, signs with other team, does poorly [winner: team] (AJ Burnett)
3.) player does well, opts out, signs with original team, does well [winner: player]
4.) player does well, opts out, signs with original team, does poorly [winner: player] (CC Sabathia)
5.) player does poorly-as expected, doesn't opt out, stays with original team, does well [winner: draw/team]
6.) player does poorly-as expected, doesn't opt out, stays with original team, does poorly-as expected [winner: draw/player]
Now, I am going to assume that we don't need to list every combination, unless someone can justify inclusions of the situations where a player does poorly and opts out (never happened, shouldn't happen if he has an agent), or where a player does well and doesn't opt out. We don't have a huge number of examples to draw on here (but I used some examples above, because only a handful of players had received this contract clause as of 2014, and while in the past year they have become a little more common, we haven't gotten to see the results of those yet.
So basically, out of the six most likely scenarios, the team wins 1.5 of them, the player wins 3.5, and 1 is a draw. Now, we can argue the probabilities of those outcomes, and if we knew that we could create an expected value matrix, but that seems to be the situation we are facing as of today. The opt out provides a likely additional benefit to the player at the time of signing, but the game theory side of things creates possibilities (even if slim) for the team to end up benefiting more in the long run, or at least transferring the downside risk to another team.