The interesting thing about https://ghola.dev/take-1/ being slow is that the slowness wasn't apparently until I thought I was done. Whatever I did to tank performance happened on a slow boil. Every time I worked on it, it seemed fast and/or fine. Until it wasn't.
What that taught me was that mark/measure is necessary, but not sufficient to understand performance, and the browser's dev tools performance tools are probably not really used widely, since there are effectively no docs