this was not optimized for processing thousands of short tasks per minute, more processing one big task like 48 bit PoW
Discussion
well it's easy enough to fan out the work just by dividing the nonce number space by the number of threads and each one then only has to increment and hash