In addition, it depends on the statement made/ conclusion withdrawn. Pinpointing a pathway or a candidate gene would indeed require validation. You might afford to skip validation in profiling experiments given that your experimental design is flawless and you use stringent statistical analyses.

Matthew, I basically agree with Sebastian and Vipul: the depth of the validation depends on what you want to say or how much "strength" your results should/must have. Do you want to quantitate? Is precision important? LLOQ? What about linearity? Ion suppression? Effectivity of digestion? If you answer "yes", then you're talking about a classical validation (FDA, EMA) meaning, indeed, every result. Even if you're just screening, the results must have a certain validity. I don't see the necessity of validating the tests with a second method if the current method, itself, is validated. It's ok to do comparisons in the scope of a validation, but as I see it, that's not absolutely necessary. Good luck!