Proxies are surprisingly fast in general, the code in V8 is relatively simple. It’s hard to say more unless I know what you want them to be compared to? I only mentioned it because I’ve found people assumed, just as I did, that “boy I bet your polyfill is slow cause Proxies likely are”.
I intend to create a deep technical dive including benchmarks but at this point the naive reference implementation is just for high level demos, so judging real numbers from it isn’t completely fair. Right now it blows the pants off React in speed, but theoretically that shouldn’t be the case — likely is due to React simply doing more work for correctness and supporting multiple paradigms (e.g. class components vs. functional). I imagine that gap would narrow if my code had the robustness React has.
Also, the idea is to have browsers implement LazyDOM so perf would be even more unpredictable; sometimes that means faster sometimes (as you mention about ES6) that means slower initially. If they did implement it, how they did it is an implementation detail and no longer related to the ES6 Proxy I use in the demos.