prettyprinter configures all benchmarks with
https://github.com/quchen/prettyprinter/blob/11e8bf12aed2f698e619c3a8c07eca3af4da1aed/prettyprinter/prettyprinter.cabal#L202
What's the reason for this? (I suspect a copy-paste from test suites?..)

I mean, it does not make sense to run benchmarks concurrently, because results would be wildly wrong. So neither criterion nor gauge nor tasty-bench do it. The only effect of such configuration is parallel GC over all cores, which in my experience (especially with otherwise default RTS parameters) negatively affects relevance of performance metrics.