Investigating the appearance of NaN in criterion's output, I found that NaNs were frequently introduced into the resample vectors when the resamples were sorted. Further investigation of the sorting code in vector-algorithms revealed no bugs there, and if the runtime was forced to keep a keen eye on the indices, by replacing unsafeRead/Write/Swap with their bounds-checked counterparts or by 'trace'ing enough of their uses, the NaNs did not appear. I could not reproduce the behaviour with ghc-7.0.1 (using exactly the same versions of the involved libraries), ghc-7.0.2 (different criterion release, the other libraries identical) or unoptimised compilation with 7.0.3 (no NaNs encountered in some 100+ testruns with varying input). So, is it possible that some change in ghc-7.0.3 vs. the previous versions caused a bad interaction between ghc-optimisations and vector fusion resulting in bad vector reads/writes?