New:
- Host disk usage is reported correctly (requires the agent to see the host filesystem: "-v /:/hostfs:ro" for the containerized agent, system service agent has access to this by default)
Fixed:
- Nightly host system refresh now works correctly
- Container stats collection stays alive across stream failures — a single failed stats stream no longer stops collection for the rest of the containers on the host.
- Host metrics (CPU/memory/disk) are routed to the server's alert evaluator, so host-scope alert rules work for agent hosts.