You’re running the non-LTS latest/stable snap channel.
For LXD this means you cannot downgrade after an upgrade (only the LTS series allow for downgrades).
For LXD clusters all cluster members need to be on the same snap revision.
So the way forward is to perform a sudo snap refresh lxd on all cluster members.
LXD has built-in logic that detects when one cluster member has been updated by snap and will then ask the other cluster members to perform a snap refresh too, so as to minimize the time between to all cluster members being on the same version.
However snap has this concept of cohorts and phasing, which can mean that depending on how the cluster was originally installed, some members may not be in the same cohort and will not see the update straight away.
Or there could be another reason why some member(s) are not being upgraded.
Either way you should perform a snap refresh on each member and then if the cluster doesn’t recover please can you get the output of the LXD daemon logs on each cluster member, as well as the output of sudo snap info lxd.
For more information on managing when the snap performs its updates see:
Thanks