HEAD 7 = 88f910772dcdf18349fd1a4d8783189e0fc9bad3 (assembly/v0.69.0; refs/gate/head7 pinned locally, pushed by OBJECT to kitsubito, rev-parse verified). Parents 379b80f1 + 9547cd49, tree 60fdf3fa. #295 both legs green: Linux 3127/3127 clippy 0 treqs 0; Windows spt+spt-daemon set + daemon_status_pid_e2e 1738/1738 (1 leaky), the e2e cell listed once in listing.raw and executed PASS 11.4 s, clippy 0, treqs 0. treqs on the head rc 0, 899 OK (+1). Whole-tree marker grep 0 files. THREE conflicts this time, all in the message: brain.rs net_status (both sides converged on the same bounded body; head's [impl->REQ-BRAIN-READ-BOUNDED-PER-CALL] tag line kept, sitting on the call_deadline line it evidences), plus CHANGELOG.md and docs-site/src/changelog.md as UNION. Unbounded read_event() census in the head-7 tree equals head 6 (tests/input_ack_deadlock.rs + xtask brainread only), so the member's brain.rs edits admit no new unbounded call site. [0.68.0] block 406dd3cdd2a37844, byte-identical to head 6; both doc tags present in [Unreleased]. ONE MORE AGAINST MYSELF, worse than the CRLF one: my first #295 resolution wrote the docs-site changelog page INTO the root CHANGELOG.md — the two conflict blobs were spooled to scratch files named c295-CHANGELOG.md.conf and c295-changelog.md.conf, which on this NTFS box are ONE file, so the second write won. The marker scan is blind to it by construction; the [0.68.0] byte-identity check caught it (sha went to 4c62c54a and both doc tags read absent). Rebuilt from distinctly named files; the committed head 7 has the checks above. Your per-head [0.68.0] check is doing real work — keep it. Head-7 compile gate (clippy / brain-read-check / xtask check / nextest xtask / treqs) is running on kitsubito now; #287b Windows leg running on the rig. #287 = head 8 when its Windows leg lands, then the scope commit.