The "z" just means "this line has 52 more bytes encoded on it". If you look carefully, the last lines in the file actually have a different marker, meaning "this line has (some specific number of bytes smaller than 52) on it" (e.g., "y" = 51, "x" = 50, something like that).

The reason it needs this is that it always produces output in 5-byte chunks, so the output for these inputs is exactly the same:

X<eof>
X\0<eof>
X\0\0<eof>
X\0\0\0<eof>

Say that output is "ABCDE". The encoder will produce something like this:

X<eof>       -> aABCDE
X\0<eof>     -> bABCDE
X\0\0<eof>   -> cABCDE
X\0\0\0<eof> -> dABCDE

The "number of bytes on this line" marker lets the decoder later figure out how many trailing NULL bytes it needs to stick on the end. This is a bit simplified, but I think it's roughly what's going on. The "z" thing is also maybe kind of a checksum to help tell if the file got damaged in transit?

Before, we were just testing base85($data), but now we're testing kind-of-format-as-a-git-patch(base85($data)) so that I can inline base85() later.

The behavior of kind-of-format-as-a-git-patch(...) is to break the data into 52-character lines, encode them, then stick "z" in front of each of them (if the line has 52 bytes worth of data) or some other character (if it has fewer), so that's why the new stuff is line-wrapped and has a bunch of "z". The actual text should otherwise be the same.