I tried the code again and verified the output size to be 199243770 which is correct. It finishes in 1.178s.
https://gist.github.com/irony/d44cbda0c71707e89274edd63fcc3d35

I do think there are few cases where an automatic threading model will be faster than a designed clustered solution but as you say- time will tell ;)

Anyway, good luck and I look forward to follow your progress, impressive stuff!