In the blog post you mentioned 14.3 and showed ifconfig commands that looked like the old system. Since switching to the new system here I've used host addresses exclusively on the vlan* members of the bridge - it wasn't even allowed to put addresses on the bridge itself. I vaguely recall a sysctl could change that.
My ipv6 network at home is a bit of a mess. I have:
* gateway box fetches IPv6 prefix-delegations with the antique WIDE dhcp6 client from 2008 and assigns things to subnets
* gateway box runs many rtadvd announcements for dynamic addressing and ULA assignments. There are both temporary and "permanent" ULA prefixes.
* gateway box runs kea dhcp server for both v4 and v6 which includes giving out fixed public and ULA ipv6 address assignments
* FreeBSD boxes (desktop, server, and bhyve VMs) have a mixture of static IPv6 addresses configured, and kea-assigned addresses, and SLAAC style self-assignments.
* an active tayga 6-to-4 translator:
(ping6 github.com) 16 bytes from 64:ff9b::8c52:7404, icmp_seq=0 hlim=50 time=33.716 ms
* bhyve VMs with tap interfaces assigned to bridges with SLAAC or set in static configs. Some with SR-IOV vf interfaces on a second interface.
* VNET jails with single or multiple epairs, most static configs but they also fetch temporary SLAAC addresses.
* there are apple devices announcing rogue prefixes for Matter/Thread radio mesh subnets.
* there's a home assistant box doing who-knows-what that is also tunneling ipv6 <-> Thread radio meshes.
* The apple devices have a nasty habit after a reboot of deciding "OH HEY LETS RENUMBER ALL THOSE ULA PREFIXES!!!! It'll be FUN!"
This setup tortures the network in every way I can imagine but there's no difficulty getting addresses with the new bridge setup. If anything there's too many addresses. But none on a bridge interface anywhere.
My ipv6 network at home is a bit of a mess. I have:
* gateway box fetches IPv6 prefix-delegations with the antique WIDE dhcp6 client from 2008 and assigns things to subnets
* gateway box runs many rtadvd announcements for dynamic addressing and ULA assignments. There are both temporary and "permanent" ULA prefixes.
* gateway box runs kea dhcp server for both v4 and v6 which includes giving out fixed public and ULA ipv6 address assignments
* FreeBSD boxes (desktop, server, and bhyve VMs) have a mixture of static IPv6 addresses configured, and kea-assigned addresses, and SLAAC style self-assignments.
* an active tayga 6-to-4 translator:
(ping6 github.com) 16 bytes from 64:ff9b::8c52:7404, icmp_seq=0 hlim=50 time=33.716 ms
* bhyve VMs with tap interfaces assigned to bridges with SLAAC or set in static configs. Some with SR-IOV vf interfaces on a second interface.
* VNET jails with single or multiple epairs, most static configs but they also fetch temporary SLAAC addresses.
* there are apple devices announcing rogue prefixes for Matter/Thread radio mesh subnets.
* there's a home assistant box doing who-knows-what that is also tunneling ipv6 <-> Thread radio meshes.
* The apple devices have a nasty habit after a reboot of deciding "OH HEY LETS RENUMBER ALL THOSE ULA PREFIXES!!!! It'll be FUN!"
This setup tortures the network in every way I can imagine but there's no difficulty getting addresses with the new bridge setup. If anything there's too many addresses. But none on a bridge interface anywhere.