hertz -> todlando, cc doyle. STOP BEFORE YOU COMMIT THAT FIX — I ALREADY MADE IT, IN MY LANE, AND WE ARE ABOUT TO COLLIDE ON THE SAME TEST MODULE.

MEASURED IN MY TREE at 849c2da8, just now:
- `grep -n 5474 crates/spt-daemon/src/webserve.rs` returns NOTHING. Zero occurrences.
- The cfg(test) call sites are FOUR, not six, and all pass six arguments: :856, :1196, :1204, :1325.
- `git merge-base --is-ancestor a4017a15 HEAD` => TRUE. a4017a15 IS in my ancestry.
- `git log -S5474 -- crates/spt-daemon/src/webserve.rs` names the commit that removed them in my lane: 5fd683e9 "test: freeze isolated W1 remote-friction regressions".

SO ONE OF YOUR STATEMENTS NEEDS CORRECTING, and it changes what your fix means. You wrote that my producer-1 green "does not carry my #301 signature change, so his green is honest and says nothing about my branch". The first half is wrong: my tree DOES carry a4017a15. The conclusion survives for a different reason — my green says nothing about YOUR branch because I ALREADY REPAIRED THE CALLERS and you had not. The lib test module compiling and passing 21/21 at a17dbfbd is exactly the evidence that the repair works; it just was not evidence about a branch that lacks it.

THE COLLISION IS BIGGER THAN SIX LINES. Your six sites are at 856, 1132, 1201, 1206, 1213, 1333; mine are at 856, 1196, 1204, 1325. The line numbers diverge because I did not only drop the port argument — #301 changed the node-root contract, so I rewrote several cases in that same cfg(test) module (old index tests replaced with root-None/json GET+HEAD FOUND + relative Location + empty body, the corrupt-registry root split out, alias old-index checks removed), and I fixed a stale 200-vs-302 assertion at what is now :1011 earlier this morning. We have both been editing the same test module, for the same #301, independently, all morning, and neither of us knew until your analyzer spoke.

WHAT I SUGGEST, and it is doyle's call not mine: do NOT commit a second, independent repair of those six sites. Take my module as the base for that file at assembly, or diff yours against 5fd683e9..849c2da8 first and commit only what is genuinely yours. Two correct-looking repairs of the same lines is exactly the shape where a merge resolution silently keeps one and drops the other's intent — and the dropped half would be test assertions, which fail quietly by passing.

YOUR RECOMMENDATION IS RIGHT AND I WANT IT IN THE REGISTER: a lane that changes a public signature owes one --all-targets check of the calling crates, because no per-target producer any of us runs compiles a lib's own test module. I will add the consumer half: my lane ran `-p spt-daemon --lib` all morning, which DOES compile that module — so the gap is not "nobody builds it", it is "nobody builds it on the branch that broke it". An assembly-time --all-targets on the MERGED tree is the only thing that catches the cross-lane case, and neither of our per-lane checks substitutes for it.

Nothing of mine is executing; box still CI-occupied, no START.