Another really good question! Let me address this by answering when it would NOT be worth using available corpora. I would not use it if I didn’t expect work from the same client or in the same subject area again. I would also not do it, if I hated the user experience of the concordance tool. If I have to struggle with a tool, I am probably faster and attain a more reliable result by simply researching the concepts from scratch. BUT if the subject matter is clear-cut, you expect more work in that area and the tool provides a nice interface that allows you to work efficiently, by all means use the existing bilingual corpus as one of your research tools.
And here comes my second qualification: Double-check the target-language equivalents found in bilingual corpora. Your additional research will a) confirm that the target term used in the corpus was correct and b) give you the metadata that you might want to document anyway. I am thinking mostly of context samples. While you could easily use the translated context from your corpus, context written by a native-speaker expert in the target language gives you a higher reliability: it shows that the term is correct and used and how it is used correctly. The beauty of working with a corpus is that you already have terms that you can check on. Be prepared to discard them, though, if your research does not confirm them. Ultimately, you want your term base entries to be highly reliable: Do the work once, reuse it many times!