The self-hosting check cannot run on this runner

The job I added failed with `env: can't execute bash`, and the reason is worse
than a missing package. This project's runner is a docker executor with
`privileged = false` and `volumes = ["/cache"]`, so the job gets an alpine
container with no bash, no docker CLI, and no path to a daemon. I wrote and
tested the script against the host shell over ssh, which is not what CI is.

Wiring it up needs a runner change — privileged, a socket mount, or a shell
executor — on a machine that also serves production. That is not a change to
make on the way past, so the job is removed rather than left failing or, worse,
made to skip. A job that always skips is the pattern I spent this afternoon
removing from the Android suite; adding one here would be indefensible.

The script stays, and says plainly where it has to run and why it is not
automatic. It has earned that much: on its first two runs it found the missing
`docker network create web` and a documented password that broke the relay
about three times in four.

⚠️ So this closes #4 only in the sense that the check now exists and works.
"Nothing boots the quick start in CI" is still true, and now has a named
blocker rather than an absence.

.gitlab-ci.yml

@@ -170,23 +170,6 @@ conformance:
   timeout: 40 minutes
 
 # What gets a package wrong is missing files, not failing logic.
-# 🔴 Nothing in this repository ran the instructions it publishes, so when the
-# quick start broke it stayed broken — three faults at once, and the last one
-# survived a day of nobody noticing. The people it fails are self-hosters, at
-# the moment they are deciding whether this project works.
-#
-# ⚠️ It BOOTS rather than validating YAML. `docker compose config` renders that
-# file happily with no `web` network in existence — the external-network check
-# happens at `up`, and so does the database URL the password goes into. A job
-# that only linted would have passed throughout both outages it exists to catch.
-selfhost:
-  <<: *rules
-  stage: test
-  tags: [vps]
-  script:
-    - Scripts/selfhost-smoke.sh
-  timeout: 15m
-
 # 🔴 The generated pages are what people actually read. A source can be fixed
 # and the page left stale, and nothing says so — /self-hosting was served for a
 # day telling self-hosters to run `docker compose up -d` without the

Scripts/selfhost-smoke.sh

@@ -16,6 +16,23 @@
 #   Scripts/selfhost-smoke.sh            # published image, the documented path
 #   Scripts/selfhost-smoke.sh --build    # build from source, the ARM path
 #
+# 🔴 **Not wired into CI, and not for want of trying.** This project's runner is
+# a docker executor with `privileged = false` and no socket mounted, so a job
+# here gets an alpine container with no bash, no docker CLI and no way to reach
+# a daemon — `env: can't execute bash` is as far as it got. Running it in CI
+# needs a runner change (privileged, a socket mount, or a shell executor), which
+# is a decision about a machine that also serves production.
+#
+# ⚠️ Until then this is a RELEASE-TIME check somebody has to run, on any machine
+# with docker:
+#
+#     ssh <box> "cd /srv/prod/messenger/src && ./Scripts/selfhost-smoke.sh"
+#
+# A check that only runs when remembered is weaker than one that cannot be
+# forgotten. It is still worth having: on its first two runs it found the
+# missing `docker network create web` AND a documented password that broke the
+# relay three times in four.
+#
 # Leaves nothing behind: a scratch directory, its own compose project name, and
 # a trap that tears down volumes as well as containers.
 set -euo pipefail