It's a tough situation. If you incentivize fixing "problem accounts", then you create the perverse incentive for people to create problems so that they can fix them and earn more.
Any incentive program needs oversight to watch for the most common abuses, which means that it needs to be simple enough to spot, and managed by people smart enough to maintain it.
I manage the incentive program for my department where I work, and I can tell you that it falls into what I feel is the 3-leg stool equation.
1. It has to benefit the customers
2. It has to benefit the employees
3. It has to benefit the company
If you can pull this off, you're good, but a BIG PART of this is human understanding.
Example. Last month one of my teams spent the entire month dealing with a messy bunch of clients from an acquisition. As such, their productivity (by the raw numbers) were way below the minimum thresholds for participation in the incentive program.
Their supervisor brought this concern to me. I'm not about to punish one of the best teams I have because they busted their asses to provide good service to clients we just gained from another company we purchased (and want to retain!!!).
So I said fine, those techs get an average of the 3 previous months' performance for bonus payouts for the month of August.
The techs were very happy with this (and continue to not shy away from work just because it's "difficult" or may detract from the raw numbers everyone is bonused on), their supervisor is the hero because he looked out for his troops, and I'm the understanding manager because I understand that no numbers for any incentive program can exist in a vacuum.
Productivity continues so the company benefits, the customers benefited and will continue to do so, and the employees benefited -- but only because human understanding made for reasonable exceptions.
If you don't run an incentive program with these kinds of approaches, you deserve the mess you inevitably get.