I don’t think that’s a fair characterization, you should be able to redefine widely-used Elisp functions to run in deno. One that immediately comes to mind is the crappy HTTP client that Emacs ships with. (I have not tried this, just speaking hypothetically.)

... That still isn't an improvement of elisp speed. It might improve the performance of Emacs, but it's not improving the performance of elisp.

Now we're just quibbling. No it doesn't improve the speed at which the Elisp interpreter can churn through Elisp bytecode. We can however dynamically improve the speed at which some Elisp functions return, which would have the net effect of speeding up consuming Elisp code.