I use a flat rate system. Or approximately flat at any rate. We charge a flat rate for system clean up/detox, and the same flat rate for a reinstall. In general, we make the judgement that a given system will clean up OK or else will require a reinstall fairly early in the process, often together with the customer when they bring the machine in and we have our first look at it. Most of the time we get that judgement right: if it looks ugly and/or the setup looks easy to replicate, we tend to recommend a reinstall; if the setup is complex and will require a lot of trouble to get right again or the apparent problems look reasonably straightforward, we tend to recommend repairing the existing install.
We retain the option to vary the flat rate by a reasonable amount if the job turns out to be particularly easy or particularly time-consuming, but exercise it only infrequently, and never vary it either way by more than about 30 or 40%. Most people are happy with this. The ones that are not tend to be unreasonabale bast*rds anyway, so I'm more than happy to send them elsewhere. 99% of our customers are nice people and easy to work with, and we do our best to spot the other 1% in advance and get rid of them before we start. ("Sorry mate, I don't like you and I think you are going to be trouble" doesn't go down well, but "sorry mate, we are very busy this week, might be able to fit you in sometime next month" works just fine.) (It helps that we are pretty busy nearly all the time anyway - saves telling untruths.)
Having made the judgement and quoted the usual flat rate, we go ahead and do the work. Mostly, we get it right, but from time to time we spend a fair while on a system before deciding that we are not going to get it right within any reasonable time-frame and have to reinstall anyway. The flat rate remains the same, we just wear the lost time. It was our judgement that the clean-up would be OK, so we wear the consequences. We aim to make enough on the standard jobs (the ones we guess right the first time) to cover us for the nasty ones. (Depends a bit on the circumstances too: if the job looks nasty but the customer is dead keen to retain the original setup and wants us to try anyway, we pre-warn them that it could cost a bit more than usual and take a bit longer. Depends on their time constraints too: if they are happy to wait a bit longer we can do it in low-priority time and keep the cost down; if they want it yesterday, we have to put off other good paying work to get it done and charge accordingly.)
