Something I've noticed is that Dead Code Elimination for Javascript sucks. Yes, there are the (relatively) single "tree shaking" algorithms, but there because of various language issues, it's very hard to mark code as completely unused. Because of this it's extremely common to pull in a library and have a bunch of code that's never used in production!
Performance on the web is more than just runtime. That time spent downloading the Javascript and parsing it is part of the page's load time, even if the Javascript isn't executed.
