Huh. I guess I didn’t dig enough into the WebKit source — don’t know that ecosystem that well. I found some source indicating it was using crypto, then I ran my tests and they performed as if they were using crypto (i.e., very slowly). If what you’re saying is true they should probably upgrade as well.
The visualization is just one test. A crappy algorithm may do well, but a good algorithm won’t do poorly. I a visualization that would highlight V8's weaknesses, so it doesn’t surprise me that it wouldn’t demonstrate Safari’s, particularly because it looks like their algo is an xorshift which is quite different from the LCG-ish algorithms used by SM and V8.