Yes, this works, but only with some limitations. Proxmox itself has no native mesh awareness, it simply uses whatever interfaces Linux provides. If you want your hosts to remain reachable when a switch fails but the mesh is still alive, you need to handle that at the network layer. The common approach is to use bonding (LACP or active-backup). You can do this with regular switch ports or with mesh links. In a true mesh setup (wireless or multiple L2 bridges), you treat the mesh as an interface and include it in a bond, giving you automatic failover. The key point is to configure Spanning Tree properly or use a mesh protocol like batman-adv or OLSR to avoid loops. Proxmox itself won’t notice the failure; it just sees a working interface. For the cluster service (corosync), you can also configure multiple links in /etc/pve/corosync.conf so that it automatically switches to the other connection when one fails.
