VMHost interface reported twice

looking at the db, these are the interfaces:

maasdb=# select * from maasserver_interface where id in (31, 50);
-[ RECORD 1 ]-------------+------------------------------
id | 50
created | 2021-04-28 15:17:06.926785+00
updated | 2021-04-28 15:17:06.926785+00
name | eth0
type | unknown
mac_address | 00:16:3e:a5:db:e7
ipv4_params | ""
ipv6_params | ""
params | ""
tags | {}
enabled | t
node_id |
vlan_id | 5001
acquired | f
mdns_discovery_state | f
neighbour_discovery_state | f
firmware_version |
product |
vendor |
interface_speed | 0
link_connected | t
link_speed | 0
numa_node_id |
sriov_max_vf | 0
-[ RECORD 2 ]-------------+------------------------------
id | 31
created | 2021-04-26 07:43:30.735627+00
updated | 2021-04-26 07:43:30.735627+00
name | eth0
type | unknown
mac_address | 00:16:3e:14:3d:56
ipv4_params | ""
ipv6_params | ""
params | ""
tags | {}
enabled | t
node_id |
vlan_id | 5002
acquired | f
mdns_discovery_state | f
neighbour_discovery_state | f
firmware_version |
product |
vendor |
interface_speed | 0
link_connected | t
link_speed | 0
numa_node_id |
sriov_max_vf | 0

Oddly, they're both "unknown" type, and MACs don't match any interface on the machine where the virsh daemon is running.

Instead, the latter is one of the MACs of another rackd registered in maas (a separate LXD container that's not currently running).

Turns out, this was caused by having the virsh host registed as "localhost", which causes confusion when you have multiple rackds.