At a glance this looks like an obviously nicer format that a zip of jpegs, but I struggle to think of a time I thought "wow CBZ is a problem here".

I didn't even realize random access is not possible, presumably because readers just support it by linear scanning or putting everything in memory at once, and comic size is peanuts compared to modern memory size.

I suppose this becomes more useful if you have multiple issues/volumes in a single archive.

Random access is completely possible within a zip, to the degree that it's needed for cbz; you might not be able to randomly access within a file, if for some reason the cbz was stored with deflate on a jpeg, but you can always access individual files independently of each other, so seeking to a random page is O(1).