D036 - Entry 0x80 is a digest manifest over named things, and two of its slots are established

Twelve 32-byte slots, six of them non-zero, and it is not a table over entries - the digest table at 0x1 already does that. It digests specific, named things:

offset	holds	samples
0x00	D2 56 01 00, then zeros, then a small value	identical in 3/3
0x40	SHA-256 of the whole filesystem image, from the header's offset to end of file	3/3
0xC0	SHA-256 of the param.sfo entry	3/3
0x20, 0x60, 0xA0	digests of something not present in the package	unknown

0x40 is the one that matters for a writer: it makes the image digest computable, and it is the thing a package is about.

Three slots are left unexplained and are reported that way by selfish derive rather than rounded up. They were checked against every entry, every region of the package file between every pair of interesting offsets, and the SHA-256 and SHA-1 of all 134 files extracted from the three packages. None matched. Whatever they cover is not in the package.

The leading four bytes are recorded as observed and not interpreted. They are identical in all three samples while every digest slot differs, so they are clearly not a digest - but "the same in three files" is an observation, not a meaning, and naming it would be the invention principle 5 forbids.

0x400, 0x401 and 0x1002 remain wholly unknown. All three are dense and high-entropy in every sample, and none contains the digest of anything tried. 0x400 is 1024 bytes and 0x401 is 512 - four and two RSA-2048 blocks respectively, which is a shape worth noting next to the fact that the two identified key entries are 2048 and 256 bytes. That is a shape and not a finding.