In part this is because I feel inline links on wikipedia often lets you overestimate how well you comprehend a subject. It's as though the fact that there is a link over a word makes you think you could go check what that is, and therefore that you probably understand the word.
Actually looking up some term mid sentence is also probably not particularly helpful for understanding the text either.
I concede that this is a somewhat unusual decision, but I designed the pages for the singular purpose of reading and comprehending the article. Everything else is secondary. It's a bit of a design experiment, but I think it's worked out pretty well. I don't fall into skim-reading these articles like I must admit I often do with regular Wikipedia articles.
I was actually looking at the footnote links, and those do indeed seem to be missing entirely -- as well as the whole references section. Arguably that's more important than links to "definition" wiki pages, so it forms the basis for the knowledge (vs mostly just definitions).
If you want to look up sources it's better to use the live wikipedia at any rate, since it will be more up to date than this snapshot.
My mirror clocks in at 21 Gb. I'm self hosting so there's no real money cost besides the one-time investment in hardware. A really chap person could probably host this off a raspberry pi attached to something like a Corsair Voyager and serve multiple requests per second.
There's some technical trickiness in actually storing the files in a file system, since most filesystems don't deal well with having millions of files in the same folder. So I've had to build a four-tier directory structure based on the hash of the file name to be able to store the files. They're stored in a structure like 31/444/781/225/foobar.gz.
I did some experiments storing them as BLOBs in a database, but I couldn't get it to work.
The bandwidth usage is pretty inconsequential since most of my pages clock in at just a few kilobytes.