2026-09-04 - (/loop) A measurement was sitting in the list of things nobody knows

745 questions / 140 premises   ->   743 / 136
two guards added, both made to fail; two plan items came back clean

Twenty-seventh cron tick, working the plan's three named items.

The item that was predicted wrong

scePthreadMutexUnlock was flagged for a suspected overclaim - its question says the correspondence is inferred "from the name and the guest's usage" while known_by said assumed. It was the reverse. The entry carried, in assumptions:

CONFIRMED ON HARDWARE: unlocking a mutex nobody holds returns 0x80020001.

questions ranks assumptions and backlog 022 is generated from that ranking, so the queue was asking a console to establish something a console had already established, in a sentence saying so in capitals, behind 4,999 calls.

Beside it, one D398 retired months ago: "Vendor error codes appear to be 0x8002_0000 | errno ... a structure worth testing for, not an established encoding." D398 provoked seven failures across five families on hardware and every one came back in that shape - and this entry's own code was one of them. A sentence that outlived its gap (check 13).

Both moved to edge_cases with their reasoning intact. known_by is now guest-observed, matching the sibling scePthreadMutexLock - an inference from 4,999 proceeding calls, not new evidence, and deliberately not measured, because two open questions remain and check 10 forbids one measured fact promoting an entry past them.

The two that came back clean, filters stated

Namesake test, widened past D540's single phrase. Ten premises claim a POSIX correspondence; the guard covered one. Widening the candidate set to several prefix strips - sceKernelWrite gives both write and kernel_write, since which prefix is the vendor's cannot be read off the name - and applying it to all five claim sentences plus the negative one:

same shape (14fn, 872,904 calls) 0    behaviour follows (3) 0    resembles (1) 0
same name (32) 0                      "there is no POSIX function" (9) 0 that do have one

Cross-references inlined. "As _open." and "Same delivery caveat as posix_sigemptyset." now say what they mean; six entries joined the premises they pointed at.

The exception that turns the D537 split into a test

Inlining put "Nothing here delivers signals, so a handler never runs" in front of the D537 test, and it is plainly a statement about orbistoun. I was about to move it. The implementation says why not: "Recorded as an assumption rather than implied by this reporting success."

edge_cases are not in the ask list. An entry whose success return is a lie needs that where its unknowns are read. So the split is a test, not a rule - what orbistoun does goes in edge_cases unless leaving it there lets a success be read as working. Reading the code before editing is what caught it.

Guards

an_open_question_does_not_announce_a_measurement (new) and a_claimed_posix_namesake_exists_and_a_denied_one_does_not (widened, both directions, and it asserts the widening still rejects the three names D540 caught). Both made to fail by reinstating the faults.

Decision: D541.