2026-08-24 - the prober reached the end of the list

reached the end of the list, after four resumed sessions and 26 rounds on the final pass.

entries	32,466
answered	32,275
`zero`	31,111 - 96.4%
`error-shaped`	961
`rejected`	125
`value`	78
never returned	191

282 of 342 libraries answered only zero - 24,532 symbols, not one refusal among them. The same loader's address census reports ~35,000 of ~35,000 present.

Every call passes nothing, so a function taking an out-pointer cannot succeed and one taking a handle cannot operate. The correct answer for nearly all of them is a refusal, and refusing is exactly what a generic stub cannot do. COMPATIBILITY.md argued presence-versus-behaviour in prose from the day it was written; it now carries the measurement, with the caveats that matter - one loader, no hardware, and a proportion rather than any individual record.

The flag sweep answered orbistoun's first question, negatively and definitely

Four queries at offset 0 with flags 0, 1, 2, 4. All returned 0, all wrote byte-identical buffers, so no first-differing-byte record was emitted:

offset 0    00 00 00 00 00 00 00 00     start  = 0
offset 8    00 00 01 00 00 00 00 00     length = 0x10000
offset 16   03                          type   = 3

shadPS4 ignores the argument. That is worth having: their guest makes 87.6 million of these calls, and anything inferred from agreeing with this emulator is inferred from a constant.

Two things noticed and not reconciled. 130-layout's own comment records "four experiments on the emulator side went into establishing that it writes 24 bytes" - this writes 17. Both are emulator observations and neither is authority. And a 64 KiB length for the first extent is not plausibly a console's.

And a correction of framing rather than fact

The facility scheme was sent to orbistoun as a discovery. OUTPUT.md has described error-shaped as "another facility, or an errno returned directly" since it was written, and classify() has hardcoded 0x8002 as the privileged one just as long. The two-tier distinction was in our own vocabulary the whole time. What the sweep produced that is new is the mapping and the errno-correctness evidence, not the idea. Said so on the bridge, because "they discovered a structure" and "they measured one their format already assumed" support different confidence levels, and they are writing to a knowledge base off these entries.