Comment by nickcoffee

The practical question I keep coming back to: if the output is meaningfully different and faster, at what point does the reimplementation argument become less about the code and more about the reputation and distribution the original project built? That seems like the harder thing to replicate, and the harder thing to protect.