Record that the CI build-gate is designed but has never run

The header claimed, in the present tense, that this workflow serializes builds
at cluster scope. The forge reports 0 recorded runs across 288 commits, so that
serialization is not in force: the local flock is doing all of it. State the
measurement and the caveat rather than let an inheritor rely on a gate that has
never fired.

.forgejo/workflows/build.yml

@@ -5,8 +5,20 @@
 # NOT safe is many agents each running `bun run build` into the shared, committed
 # web-root (public/): concurrent builds clobber each other's output and race the
 # `git add public/` that follows. internal/build-lock.sh makes that STRUCTURALLY
-# impossible on ONE machine (an flock mutex); this workflow does the same thing at
-# CLUSTER scope — it makes CI the ONE-AND-ONLY builder, and serializes it.
+# impossible on ONE machine (an flock mutex); this workflow is DESIGNED to do the
+# same at CLUSTER scope — to make CI the ONE-AND-ONLY builder, and serialize it.
+#
+# ★ STATUS: DESIGNED, NOT YET ACTIVE — measured 2026-09-22, not assumed.
+# The forge reports has_actions=true, but
+#   GET /api/v1/repos/doxa-life/doxa-maps/actions/tasks  ->  total_count: 0
+# Across 288 commits (138 of them pushed on this branch) this workflow has NEVER
+# RUN. The cluster-scope serialization described above is therefore NOT in force
+# today: every real build is serialized only by the local flock in
+# internal/build-lock.sh. Do not rely on this file to prevent a concurrent-build
+# race until a runner is registered AND a run is observed in the Actions tab.
+# (The actions/runners endpoints 404 on this Forgejo version, so runner state was
+# not directly measurable — the 0-runs figure is the load-bearing fact, and it
+# holds regardless: 138 pushes with a working runner would have produced runs.)
 #
 # THE CONTRACT for a swarm agent:
 #   edit files  ->  git commit  ->  git push   (DONE. never run `bun run build`.)