SUBS-006: POST git-remote push aborts NOBRANCH on a clean nested sub (LANDED)

Now: DONE

MEDIUM — LANDED 414d9f96 (2026-06-06). A clean detached nested sub forwarded be post -q ?/<sub>/.<parent>/<branch> with no msg/patch rows, hitting POSTPromote's missing-target arm which unconditionally returned NOBRANCH, aborting the whole git-peer push (157). A dirty detached sub instead routes through POSTCommit (which auto-creates the synthetic branch at the pin), so anything reaching the missing-branch arm from a detached wt is by definition a clean sub with nothing to promote. Fix sniff/POST.c: detached wt → POSTNONE (no-op, -q-swallowed); an attached wt still refuses NOBRANCH (DIS-020 preserved). Added test/post/29 repro.

MEDIUM. A be post ssh://<host>/<parent>.git?master (git-peer remote) recurses the local-commit step into mounted subs, but when it reaches a clean, detached nested sub whose synthetic branch (?/<sub>/.<parent>/<branch>) does not yet exist, sniff fails ?/<sub>/.<parent>/<branch> does not exist — be put ?<branch> first (NOBRANCH), which propagates as be post: aborting parent commit — sub recursion failed (exit 157). A clean nested sub should be a no-op, not abort the whole push.

Repro

Depth-3 forest (parent → vendor/sub → vendor/sub/.../leaf), leaf clean.
Make a local parent commit.
be post ssh://localhost/<rel>/parent.git?master.
Recursion descends to the clean leaf, forwards post ?/leaf/.sub/.parent/master; the branch is absent → NOBRANCH → 157.

Expected

Per spec: "be post on a detached sub auto-creates that branch at the pin" and a clean sub with nothing to commit should no-op. The remote-push recursion must not refuse on a clean sub merely because the synthetic branch has not been materialised yet.

Actual

Clean nested sub aborts the parent push with NOBRANCH. (Note: for a beagle/file: peer the push itself recurses; the .git suffix here routes it as a git peer, so only the local-commit fall-through runs — that fall-through is what trips.)

Fix sketch

In the post recursion's clean-sub branch, treat a missing synthetic branch on a clean detached sub as a no-op (skip), or auto-create it at the pin before the push as the spec promises — rather than surfacing NOBRANCH. Distinguish "clean sub, nothing to push" from "dirty sub, branch genuinely required".

Tests

Covered by test/all/01-subs-all-verbs (remote-push leg) and the post depth-3 path; a dedicated test/post case can be added once SUBS-001 unblocks the post-order commit path.