First 1024 bytes are preamble; no real text.
81 - 516 are all ff. Then 517-520 is "2 0 0 0", then "3 0 0 0",
then "q 0 0 0", "fe ff ff ff", "r 0 0 0". (Note we are missing q and r
below, at 961.)
Then we continue counting up from 537 = "7 0 0 0", to 881 is "93 0 0 0".
Then the expected "94 0 0 0" is replaced by "254 255 255 255".
Then it continues from 95 ("_") at 889, to 112 ("p") at 957.
Then again, we get fe ff ff ff instead of xx 00 00 00 for q and r,
then "s 0 0 0 t 0 0 0" as expected, then another fe ff ff ff instead of u,
then on with wxyz{|   (119-124).
Then another fe ff ff ff, going to 1013, which is ff all the way to 1024.

So the first 1024 bytes are basically just showing off that we can count!

----

1025 starts with "Root Entry" - as usual, R0o0t0 0E0n0t0r0y0.
1089 has 22,0,5,0,ff,ff,ff,ff,ff,ff,ff,ff,3,0,0,0,1,18,2,0
Then it's 192 at 1114, 70 at 1120, then the first diff:

---
First diff is 6 bytes at 1133 (in a sea of zeroes): 
ah	ay
208	192
13	154
138	20
26	234
60	107
140	143

Everything else from 1025 has been a zero.

1145 is 64 "@".

1153 is "Objects" (again, spaced with 0s)

1217 is 16,0,1,0, then 12xff.

1253 is the second diff, another 5-liner, very similar to before:
208	224	(208 was there before. 192/224 are both "round" numbers)
156	66
135	6
26	234	(as before)
60	107	(as before)
140	143	(as before)

Common 199,1 at 1259.

1261 is the next diff, identical to the first one: 
208	192
13	154
138	20
26	234
60	107
140	143

This again is followed by "199,1".

More zeroes,then "Quill" at 1281.

1345 has "12 0 1 1 4 0 0 0 255 255 255 255 12"

1381 is the fourth diff, identical to the second.

Then "199,1" again, at 1387.

Then 1389 is the fifth diff, identical to the first and third.

Then "199,1" at the end, again.

1409 is "Escher".

1473 is "14 0 1 1 2 0 0 0 5 0 0 0 11"

1509 (another 120 bytes down, again) is the next (pair of!) diffs.

Okay, I've taken this further in a spreadsheet. There's really very
little difference, AFAICT.

Still, all the stuff between the 0's and FF's must mean something!

Interesting that the first kb is "special". Would be good to find some
more "boundaries".