Fixed
Gluetun-style netns sidecars no longer fail to recreate when the VPN container is updated in the same batch (#2, @famewolf).
Updating gluetun and its sidecars (network_mode: container:gluetun) together recreated gluetun first — giving it a new container ID — then the sidecars' recreate failed with joining network namespace of container: No such container: <old-id>, because their stored NetworkMode=container:<id> still pointed at the dead old gluetun.
Docksentry now snapshots each updating container's netns owner by name before anything is recreated, and rebuilds the sidecar against container:<name> (stable across the owner's recreate) instead of the volatile ID. Resolved per-container from the live owner, so it's correct even for chained / non-head netns sharing and needs no Container Group. Applies to both the manual "Update all" and scheduled auto-update paths.