diff --git a/hermes_cli/gateway.py b/hermes_cli/gateway.py
index 361441073d..1e588e600b 100644
--- a/hermes_cli/gateway.py
+++ b/hermes_cli/gateway.py
@@ -2064,6 +2064,26 @@ def _ensure_user_systemd_env() -> None:
             os.environ["DBUS_SESSION_BUS_ADDRESS"] = f"unix:path={bus_path}"
 
 
+def _adopt_user_bus_when_started_by_systemd() -> None:
+    """Point a systemd-launched gateway at its own user bus before it boots (#104893).
+
+    A *system*-level unit (``/etc/systemd/system``, ``User=<someone>``) is exec'd with
+    neither ``XDG_RUNTIME_DIR`` nor ``DBUS_SESSION_BUS_ADDRESS``, and a process environment
+    is fixed at exec time — so ``systemd-run --user`` fails for the whole lifetime of that
+    gateway even once ``/run/user/<uid>/bus`` is up.  Every restart-safe worker crosses that
+    seam (``tools.process_registry.restart_safe_gateway_child_argv``) and it fails closed by
+    design, so cron and Kanban dispatch died at every fire on headless service installs.
+    ``_ensure_user_systemd_env`` derives both values from our own uid and adopts them only
+    when the runtime dir is really ours and the socket really exists, so a host with no user
+    manager keeps its honest "unavailable" verdict instead of a fabricated bus address.
+    Doing it here rather than at the dispatch seam matters: worker environments are snapshots
+    of ``os.environ`` taken at different points, so the adoption has to precede all of them.
+    """
+    if os.name != "posix" or not os.environ.get("INVOCATION_ID"):
+        return
+    _ensure_user_systemd_env()
+
+
 def _wait_for_user_dbus_socket(timeout: float = 3.0) -> bool:
     """Poll up to ``timeout`` s for a user systemd control socket (user@.service takes a moment after enable-linger)."""
     deadline = time.monotonic() + timeout
@@ -4530,6 +4550,8 @@ def run_gateway(verbose: int = 0, quiet: bool = False, replace: bool = False, fo
     if _absorb:
         _absorb_windows_console_controls()
 
+    _adopt_user_bus_when_started_by_systemd()
+
     # Refresh the systemd unit on every boot so restart settings stay current even after an
     # exit-code-75 respawn (stale-code or /restart), which bypasses `hermes gateway restart`.
     if supports_systemd_services():
diff --git a/tools/process_registry.py b/tools/process_registry.py
index 9580eb2a54..b51b0404c8 100644
--- a/tools/process_registry.py
+++ b/tools/process_registry.py
@@ -235,9 +235,15 @@ def restart_safe_gateway_child_argv(
     if not _is_supervised_gateway_process() or not os.environ.get("INVOCATION_ID"):
         return command
     if not _systemd_run_user_scope_available():
+        # This text is what the operator actually sees: it is stored as the cron
+        # execution's error and printed on the job row, so it names the remedy
+        # rather than only the symptom.
         raise RuntimeError(
             "cannot create restart-safe systemd scope for gateway child: "
-            "systemd-run --user --scope is unavailable"
+            "systemd-run --user --scope is unavailable — this gateway has no reachable "
+            f"user D-Bus session (expected /run/user/{os.getuid()}/bus). On a system-level "  # windows-footgun: ok — behind the _IS_LINUX return above
+            "service install, run `sudo loginctl enable-linger <gateway-user>` and restart "
+            "the gateway."
         )
     scoped = _build_systemd_scope_argv(command, unit_suffix=unit_suffix)
     if scoped == command: