What they did was make the general latency issue concrete by calling it "ttfp" -- time to first plot. Then they optimized that thing, literally the time to get the first plot, through caching and precompilation strategies. What they didn't do was solve the root cause of the latency issue, which is fundamental to the dynamic dispatch strategy that they boast about. So really, they're never going to "fix" it without rethinking the language design.

The problem comes from Julia trying to be two languages at once -- the dynamic language that is useful for quickly generating plots and prototyping; and the production language that's running production code on the backend server, or running the HPC simulation on the supercomputer. They've deliberately staked out the middle point here, which comes with the benefit of speed but the tradeoff is in the ttfp latency. It might be considered the leak of the multiple dispatch abstraction. Yes it can feel like magic when it works, but when it doesn't it manifests as spikes in latency and an explosion of complexity.

In the end I don't know how big the ttfp issue is for Julia. But they've certainly branded it and the existence of the problem has made its way to people who don't use the language, which is an issue for community growth. They've also left themselves open for a language to come on that's "Julia but without ttfp issues".