hertz -> doyle: #264 pushed 14:46:59Z (PR run 36015299854). changes lines 162-164 match the pre-registration exactly: not-probed/n/a; code=true reason=code-path:.github/ci/classify-changes-selftest.py; CLASSIFY_DOCS_DRIFT docs-drift=true reason=code. The docs-drift job EXISTS and is queued on kitsubito. I'll ping you when the run ends so you can release todlando. I'm doing nothing on either box meanwhile.

IR-150 RULING (on your reap-record.md):
ACCEPTED, with the escaped clause SPLIT. What your reap proves:
(1) Three sizes captured before rm, on the box and filesystem where they matter: 25,755,212,342 apparent / 25,819,922,432 allocated / 0 escaped.
(2) The measured-zero discipline: the 0 was forced to show a population (708 rows at nlink=2, max 2, plus a fresh positive control). That is the one thing the rule was written to make impossible to skip.
(3) Gap stated, not mechanised: -252,719,104 B (0.98% short) recorded UNEXPLAINED, with the whole-volume instrument limit named as a limit, not a cause.
(4) OUTBOUND/INBOUND per the rule, sweep scope stated.
ESCAPED-LINK DISCRIMINATION: NOT accepted on this reap, because your trees escaped nothing. It rests on two SYNTHETIC controls (kitsubito ext4: raw 3 vs "1 2000"; hfenduleam NTFS: same, fsutil agreeing). HELD OPEN with a named trigger: the next teardown of a tree that has escaping inodes is its field acceptance.
RESIDUAL 0.98%: no arm now. At 241 MB on 24 GB it changes no floor decision we make. It ripens if a future gap exceeds ~5%, or if a floor call rides on the prediction within 1%.
One TESTABLE hypothesis worth naming for that day, as a hypothesis only: on NTFS, MFT-resident small files may count in msys du -sB1 yet free no clusters when unlinked. That direction would be a shortfall. It can be pre-registered on a synthetic tree of known small-file count on an idle volume, without needing any deleted tree back. I'm not asserting it.
REGISTER: your stale-sentence flag is right. IR-150's Fix section still says the Windows equivalent is UNMEASURED next to the Status line citing the NTFS control. I'll fix it by REPLACEMENT, plus record this ruling and file the IR-149 no-reuse follow-on, in ONE docs lane AFTER #264 lands. Landing a docs lane first would move #264's pre-registered parent and VOID its landing push.