SUPERSEDED 2026-08-30 by decide-does-track-r-work-on-master-like-every-other-lane. This ticket and its sibling asked one question, and both priced it off a divergence measurement that has since moved: git cherry origin/master origin/rust reports 122 of the 136 commits already have patch-equivalents on master, only 14 do not, and 14 is an upper bound — two spot-checks of those 14 are demonstrably present on master. Do not act on the numbers below.

Filed in rejected/ because it is not the open question, not because its reasoning is wrong — the analysis below is preserved intact and is still the best writeup of the topology argument. Read it; re-measure before quoting it.

Should the Rust topic branch be retired onto master?

Filed 2026-08-29 by frank-coordinator on frank-rust's argument. The evidence is frank-rust's; the measurement and the Makefile finding are mine. I recommend retiring it, and it is the owner's topology call, not mine.

The current rule and why it exists

session-roster.md, "2026-08-27 — PER-TOPIC TREE TOPOLOGY": a checkout per role, with topic branches carrying the destabilizing targets (wasm, rust) while master is the trunk where A, B and the coordinator meet. The mechanical reason it does not extend to A/B is good and unchanged: stable_linux_amd64/** is 28 MB of committed binary that A rewrites at every pin, so a long-lived branch pays a binary conflict on it every merge.

CLAUDE.md still says all tracks work on master; the roster section is newer and supersedes it for these two trees. That divergence is itself worth settling — an agent reading CLAUDE.md alone concludes its own tree is mis-checked-out, which is what happened to me this session.

The argument for retiring it

frank-rust's, and it engages the criterion rather than preferring an outcome: the work has not been destabilizing. Every commit is compiler/rparser.inc plus tests, no new AST node, no IR op, no symtab field, no backend edit; the self-host fixedpoint converged at each one; the Rust suite is gated. Twice a change looked like it needed a Track A ticket and did not — struct/enum returns and the &mut aliasing switch were both already in the shared machinery and simply unused by the Rust frontend.

The measured cost of keeping it

Track T sweeps origin/master and nothing else. 8 commits have never been near the matrix — no cross-target jobs, no corpus. A per-commit self-host fixedpoint proves the compiler reproduces itself at the default -O level; it is not evidence that thirteen green test_rust_*.rs hold on i386 or aarch64.
Drift is already live. origin/rust is 8 ahead and 57 behind origin/master, four hours after a merge-in. The comparison case is origin/wasm at 76 ahead and 312 behind, where every finding that lane files describes the tree as it stood 312 commits ago.

The cost of retiring it, which is not zero

origin/rust touches Makefile, +38 lines — not only rparser.inc and tests. frankB holds Makefile right now for the assertion-conversion campaign (498 landed, ~1630 in flight). So the merge must be sequenced behind frankB either way; this is a scheduling constraint, not an argument against.

The three options

Retire the branch; Track R works on master like A, B, C, N and D. Restores CLAUDE.md's rule, puts R's work in T's matrix immediately, removes the merge ceremony. Costs: R's destabilizing work, if any arrives later, lands on the trunk — mitigated by the same working/ ticket lock every other lane uses, and by the fact that R's files are disjoint from every other lane's.
Keep the branch, adopt a mandatory merge-in cadence — git merge origin/master at every rung boundary, not once per session. Preserves the owner's topology. Costs: the cadence is a discipline nobody can enforce, and wasm at 312 behind is what happens when it lapses.
Keep it and accept the drift, merging on occasion as originally intended. Honest, and the status quo. Costs: R's work stays outside the matrix indefinitely.

Recommendation: option 1, with the merge sequenced after frankB's tranche two. The criterion the topology was built on — destabilizing work — does not describe what Track R has actually produced, and the branch's only measured effect so far is to keep 8 green commits out of the test matrix.

Not in scope: wasm. It is 76 ahead and 312 behind with shared-file arms already ledgered in feature-a-merge-the-wasm-branch-the-shared-file-arms, and nothing there is pre-approved. Branch permission is not merge permission, for either branch.