This is relevant for CPU bound loads. Lots of lambdas do networking stuff where 90+% of the time is waiting for an HTTP download.