On the one hand, of course you are right. But we are discussing a bruteforce program here. For bruteforce, the only indicator of the quality of work is important - it is the number of keys/addresses per second.

If we are talking about randomness, then it is obvious that the program can select starting points in such a way that one target key is far from them, and the other is close, but this is not at all an indicator of the quality of work. For example, it is possible that someone wrote a program that instantly solved a some key in a wide range, due to the fact that one of the starting points coincided with the key - will this be an indicator of the speed of work?

Personally, I'm interested in testing different algorithms, but in this case your program is not available, you only show some working logs.
I also cannot fully test the modification from the sp_, since only executables for Windows are available, and 99% of my GPUs on Ubuntu. However, I saw a slight increase in speed on Windows machines, so in my opinion this modification at least has the right to life.