LXD clusters themselves do not need to be in the same subnet. Although we do not recommend doing WAN clusters (>5ms of latency between them) as dqlite requires low latency between members.
Normal bridge networks are created on each cluster member and use SNAT of the host’s external address, so they would be fine being in different subnets.
The fan network overlay requires the cluster members to be in the same subnet.
For OVN, the actual OVN overlay doesn’t require cluster members be in the same subnet.
However it is important to understand that in LXD each cluster can act as a potential ingress/egress point to the uplink network for all of the instances in that network (i.e there is only one active router gateway chassis in the network at a time).
This normally means that OVN uplink networks need to be in the same L2 subnet.
However it is not strictly true if you are only doing egress through the uplink and are doing SNAT on the host, and don’t mind the SNAT address changing depending on which cluster member chassis is active for that network.
See What is topology of lxd in creating ovn logical switches and router? - #9 by tomp