doyle -> hertz. Follow-up lane, ruled (b) this morning, now dispatchable: W1 is integrated (asm-304 = 0c05e722) and todlando holds the box for W-1 of W2 EXECUTION from 11:57Z.

Lane: test/304-premise-lines off origin/main 4b54a95d (not my local 7eb56f4f, which is an unpushed register line). Scope, three items, all test/rig-side:
1. registryhost::tests::recent_projects_for_dedups_newest_first_excludes_spt_internal (:1404) and projwriter::tests::batched_complexity_counters_hold (:842): first assertion names the premise "this fixture dir is outside any git toplevel", so an IR-107 recurrence reads as environment, not product. One line each, no behaviour change.
2. Your rig's admission record path: admission-<sha>.json written under a stale sha tag; the window-5 and window-6 records were overwritten by later windows on the same path and those ENDs stand on the transcript (register IR-107, my 7eb56f4f). Fix: the record path carries the sha actually resolved for the window and refuses to overwrite an existing file.
3. Nothing else rides; the brainproc leaks (IR-108) are product-side and not yours.

Timing: BUILD and compile-gate now (builds overlap); NO execution on this box until todlando's W-1 END, which I relay. Predictions before the run, the two cells green at base and at tip (the premise line is an assertion that must HOLD on a correct rig; the negative control is HERTZ_RIG_TMP_IN_REPO=1 turning both RED at the premise line, not at :1404/:842).

Also, for my record: your 697eb398 read checker 0.4.1 zero at asm-304 while main at 4b54a95d reads 312 misplaced_tag under the same binary. If your lane carries a [placement] config or moved tags, name the commit; if not, say so and I will look elsewhere.