I found this approach interesting:
The way I deal with these discounts is that I have an internal target for hourly payment. For each job, I quickly fill in a small spreadsheet with the counts in the various match categories. For each category, the spreadsheet shows me how much time I should spend on them. I start doing all the discounted segments. That way, I can easily track that I don't spend significantly more time on them than what they pay for. It means recycled segments may not be optimal when finished, but I won't let outright errors pass.
The interesting part is for me in the last sentence. Whenever I have to deal with partially translated projects, possibly with legacy TMs (and if so, they were built over the years by several translators with different styles and wordings), then I find CafeTran's feature to simultaneously find and replace both in the current project and in the assigned TMs extremely handy to correct misspelled words and incorrect terms on the fly: correct them in the opened segment and fix them automatically in the rest of the project and in the TM:
And when further down in the project a Fuzzy Match is suggested or inserted, the typo or wrong term has already been corrected. I find this very time saving. I wouldn't want to miss it anymore. In fact, it's for me the only acceptable way to build upon somebody else's work.
[Edited at 2016-07-29 07:07 GMT]