# hertz docs-lane banked items (2026-09-24, from doyle msgs AG6YEJB4, XXSREV4W, QQQCDZ5G)

## IR-150 escaped-term tally (XXSREV4W, SETTLED)
- todlando W4 retire 15:40Z: zero UNFORCED (population not captured). Record: .spt/preserved/331/todlando-w5/w4-retire-ir150-report.txt
- todlando W5 retire 16:53Z: zero FORCED, 15,757 files nlink>1 / 7,878 inodes / max nlink 3. Record: .spt/preserved/331/todlando-w5/w5-retire-ir150-record.txt — header says TRANSCRIBED FROM MESSAGE, not a saved capture. Cite as labelled transcription.
- forced: doyle (708 @ nlink 2), hertz Tier 2 (178/150/326), hertz rig (170). Escaped term still UNTESTED ARITHMETIC: no sample holds an inode linked from outside its tree.
- One sentence for entry: a capture living only in a peer message + commune is not a capture; a context clear takes the raw output with it (mechanism behind the two-records split).
## IR-154: own zero UNFORCED (settled).
## IR-153 (QQQCDZ5G): 17 other-agent backslash records REPAIRED repo-wide on hfenduleam by doyle. Predicate controlled both ways (17 positive via grep -F hex-escaped backslash on .git/worktrees/*/gitdir; w6-divulge negative). repair on 17 at once rc 0, 17 "repair: gitdir incorrect"; per-tree gate 17/17 at 0 after; repo-wide 17 -> 0. Refinement: batched call's rc 1 came from 38 stderr lines on records it declined; naming only malformed records exits 0 — rc reports the sweep, not targets. Status -> resolved repo-wide; historical finding unchanged, no AGENTS.md rule needed.
