Trim of the salvaged #123106 commits (@liuhao1024) to the sweep bar: shorter
rationale comments at the three call sites, one-line fold in the messaging
route, and the helper test now also pins the STANDALONE record shape
(`served_profiles: []`, flat keys only) that #123869 reports — the same
re-key through the multiplexer rung projected `{}` there too.
Why one fix covers both issues (and the launchd/bootstrap variant on #123869):
`profile_platforms_from_multiplexer` kept only `<profile>:` keys, but a
multiplex host's primary map is always `default` (run.py `_primary_profile_name`)
and a standalone gateway writes only flat keys, so the one profile the record
never prefixes projected as empty. Both dashboard callers now fold on the
resolved profile NAME (`profile_name_for_home`) instead of a directory
basename, so `own_home.name == ".hermes"` no longer defeats the fold on the
Messaging card. Secondary projection is unchanged: a served profile still sees
only its `<name>:` keys plus the shared listener mirrors.
Not addressed here (different mechanism, tracked on #124029/#123109): the
inline `-c` bootstrap argv making rung 3 (`get_runtime_status_running_pid`)
return None; with this fix that fall-through no longer changes the projection.