Hi @Jan, thanks for reporting this. This is unexpected. Could you give us a bit more information about your project (Ratio of C/C++/Assembly, custom tasks used, generated source set used, etc.)? If you could submit a build scan, it would also be helpful.
Could you also try 4.2 RC2? Around 4.0, ParallelizableTask was removed in favor of the worker API. The native tasks were refactored to use the worker API in 4.2 and you should have the same performance and maybe a bit more as reported by some of our partners.
Thanks,
Daniel