It's quite hard to follow without the associated talk. I do recall past threads where JavaScript natives were rewritten in JavaScript and shown to be faster,

It's quite hard to follow without the associated talk. I do recall past threads where JavaScript natives were rewritten in JavaScript and shown to be faster, then they were shown to be non-standard because of a heap of weird edge cases that have to be supported correctly. It hadn't yet been shown that a fully standards compliant implementation would be faster. I think a huge amount rests on slide 8 "No JS/C++ transition cost". It should probably be the subject of significant analysis on it's own. The calulation is min(Ctime + TransitionCost, JStime). Knowing all of these values individually would give a better picture. Is it better to Minimize the transition cost or eliminate it? It depends on how much less Ctime is compared to JStime.